Government Executive - Tech

‘It would be insane’ for spy agencies to not have AI model early access, lawmaker says

David DiMolfetta — Tue, 12 May 2026 15:40:39 -0400

Rep. Jim Himes, D-Conn., the top Democrat on the House Intelligence Committee, said Tuesday it would be “insane” for U.S. intelligence agencies to not have early access to advanced artificial intelligence models that could be used for hacking and cyberdefense.

His remarks, delivered on a panel at Politico’s Security Summit, come as the Trump administration is reportedly considering a major AI executive order and debating whether the Commerce Department or intelligence community should oversee evaluations of AI models. They also come as President Donald Trump makes a planned trip to China this week, where he is expected to discuss AI matters with Chinese President Xi Jinping.

“Making sure that, in particular, where our real computational brains are, the National Security Agency, making sure they have access to the most capable hacking tools … it would be insane not to do that, right?” he said.

The NSA, the spy community’s premiere hacking, codebreaking and foreign eavesdropping giant, has been testing Mythos, a major Anthropic model that’s been held back from full public release due to its substantial cyber capabilities, multiple people familiar with the matter said.

On Monday, The Washington Post reported the Trump administration is split over whether to give spy agencies or the Commerce Department dibs at evaluating models. Commerce officials are pushing back against a White House proposal to house an AI evaluation center within the intelligence community, according to the report.

Himes said the Commerce Department should also have a role to play in AI policy.

“Across the government, we should be looking at these capabilities,” he said, adding that “we ought to be cultivating — not damaging — our relationship with the producer of this remarkable new technology,” in a nod to the ongoing legal complaints Anthropic has lodged at the Defense Department, which deemed it a supply chain risk earlier this year after the company said it wouldn’t meet certain Pentagon demands.

Himes said he doesn’t think the legal spat between the DOD and Anthropic has set back the intelligence community in the near term, though “if this drags out, if [Defense Secretary] Pete Hegseth gets a bee in his bonnet about this and just decides to target because his ego is damaged … that will be a massive liability for United States national security.”

Officials are circulating draft policy documents with language clarifying the government’s ability to use private sector tech without outside stipulations, Nextgov/FCW reported last week. It’s not clear if the contracting language is part of a coming executive order or a separate policy initiative.

The ongoing discussions highlight how the Trump administration is closely examining cyber threats brought on by advanced AI models and is looking to take a more hands-on approach toward the AI sector, despite prior laissez-faire positions.

]]>

Anthropic, Code for America pilot AI tools for SNAP eligibility support

Kaitlyn Levinson — Mon, 11 May 2026 17:26:20 -0400

The civic tech nonprofit Code for America is partnering with artificial intelligence company Anthropic to develop tools aimed at helping caseworkers enhance public benefits administration across the nation.

The organizations are working together to develop an AI-enabled solution to improve the accuracy and timeliness of benefits service delivery under the Supplemental Nutrition Assistance Program, Jana Rhyu, vice president of product at Code for America, announced Friday at a summit hosted by the organization in Chicago last week.

The SNAP Policy Navigator tool is built on federal regulations, state manual selections, official policy directives and other documents to help caseworkers “quickly and accurately get an answer to [a] very specific policy question” when they are working with clients, said Michael Lai, who leads state and local government AI at Anthropic.

The tool leverages Anthropic’s Claude chatbot and is built on a model context protocol to ensure a secure two-way connection between data sources and AI applications, Rhyu said. A caseworker can input a simple, policy-based question, such as how a client’s change in income or a new federal policy could impact their benefits, and the tool outputs an up-to-date response in plain language with cited sources and suggested next steps.

The user “gets clarity on policy, not a decision on overall eligibility. The decision stays with [them],” she said.

The announcement comes as state and local public benefit agencies scramble to comply with rule changes to the federal food assistance program made last July under President Donald Trump’s “Big, Beautiful Bill.” The law subjects participants to expanded work requirements, shifts administrative costs to states based on their SNAP payment error rates, and requires that the Thrifty Food Plan — the model used to calculate the lowest-cost nutritional meal for a family of four — be cost-neutral to changes in food prices.

Since its passage, SNAP participation has declined by more than 3 million people across 36 states as of January, and further reductions are expected once the new rules are fully implemented, according to the Center on Budget and Policy Priorities.

“Policy is constantly changing, and the complexities of policy implementation are immense, which places an even bigger burden on the caseworkers,” Rhyu said.

That’s why it’s critical for resources like the SNAP Policy Navigator tool to help reduce caseworkers’ administrative burden of sifting through and trying to apply intricate policies to individual cases, she said.

Indeed, the complex rules regarding eligibility and exemptions present common barriers to benefits access and having to explain those to residents who depend on the timely and accurate delivery of public assistance to meet their everyday needs only adds to it, Lai said.

Such challenges are exacerbated by funding uncertainty, workforce shortages and increasing caseloads that many states and localities are grappling with across the U.S.

He pointed to one former caseworker who described their job as “an email inbox that's always full, where each one requires care and attention, but you're constantly getting interrupted as you try to work through the never ending inbox of people to help.”

In addition to the SNAP Policy Navigator, Code for America and Anthropic will develop a suite of Claude-based tools to further assist benefit workers with answering policy questions, reviewing eligibility documents and drafting communications to benefit recipients, Code for America leaders said in an announcement last week.

“We know that caseworkers are really overburdened in general, but especially at this moment with HR 1 as well, and so AI shouldn't be used for AI’s sake,” Lai said. “We want it ultimately to be helping in this human way and trying to make benefits administration more efficient, more accurate and more human centered.”

]]>

Pentagon will ‘never again’ rely on a single AI provider, official says

Alexandra Kelley — Fri, 08 May 2026 16:27:35 -0400

Leadership at the Pentagon reiterated the agency’s commitment to diversifying its artificial intelligence service providers, with Under Secretary of Defense for Research and Engineering Emil Michael taking the stage Thursday at an event in Washington, D.C., to stress that his department is never being “single-threaded with any one model.”

Speaking during the Special Competitive Studies Project’s AI+ Expo event, Michael said that the recent deals between eight leading AI developers and the Department of Defense are both a private sector statement of support for working with the government, as well as a step towards the Pentagon’s goal to diversify its tech stack with different providers.

“We were single-threaded on one vendor, one AI vendor at the Department of War, and to integrate into classified systems is not just putting your software on a public cloud and having it work,” Michael said, referring to his agency’s contract with Anthropic. “These are sophisticated, protective systems that take a lot of work to integrate on, so it wasn't like I could just turn on a few other models that easily. But never again we’ll be single-threaded with any one model.”

Michael continued to say that the new deals with Amazon Web Services, Google, Microsoft, NVIDIA, OpenAI, Reflection, Oracle and SpaceX are “a statement by the biggest tech companies in the world who are involved in the AI space … and have them say, ‘We support the Department of War, we support the U.S. government, and we support the… armed services for all lawful use cases.”

Michael’s comments come in the midst of an ongoing dispute between Anthropic and the Department of Defense following the company’s refusal to have its technology used in operations involving autonomous weaponry and American surveillance.

The fallout of that dispute resulted in the Pentagon designating Anthropic a supply chain risk and the White House ordering agencies to begin removing the company's products from their tech stacks. A judge put a hold on those actions in late March pending ongoing litigation over the government’s actions.

The release of Anthropic’s advanced cybersecurity-focused model, Mythos Preview, changed the discussion. Access to Mythos and its advanced capabilities for detecting cybersecurity flaws is tantalizing for the U.S. government, prompting internal drafts of policy plans that would enable some agencies to use Anthropic’s cutting-edge model.

Michael said that the advent of Mythos signals the forthcoming evolution of cyber-capable AI models.

“The Mythos moment is really a cyber moment, and it's: ‘How is the U.S. government going to deal with cyber?’” Michael said.

Major tech companies are responding to Michael’s drive to diversify the Pentagon’s vendor portfolio. Rand Waldron, the vice president of the Global Government Sector for Oracle Cloud Infrastructure, told Nextgov/FCW that Defense officials are asking cloud service providers like Oracle to prioritize interconnectedness in the effort to avoid vendor lock-in.

“From what I can see, the Department of War has some very savvy people who … don't want to go all in on one [model] because then six months later, they may need to go all in on another,” Waldron said.

He explained that there will likely be models that are more finely-tuned to particular use cases, such as code generation, data analytics, supply chain management or targeting in warfighter operations. One model from a single provider may not effectively serve each of these workflows.

“I don't believe that all those different use cases will end up being the exact same model at any given time,” Waldron said.

The Pentagon’s desire to expand the service offerings available for its workforce has precedent. Waldron said that DOD and the intelligence community have laid the foundation for a flexible approach to AI services acquisition, citing the creation of the Commercial Cloud Enterprise and Joint Warfighting Cloud Capability contracting vehicles as the blueprints for future contracting structure.

“It's not like they're trying to replace Anthropic with another model provider,” Waldron said. “They want to replace Anthropic with four model providers.”

]]>

Inside the effort to connect Congress with the feds enacting its policies

Natalie Alms — Fri, 08 May 2026 13:30:33 -0400

Congress is flying blind on the effectiveness of the laws it creates.

That’s the thesis of a new project released last week by the POPVOX Foundation, a nonpartisan nonprofit, based on the input of 50 federal employees pushed out of their government jobs last year. The intent of this work was partly to gather insights on how policy implementation works in the executive branch and what barriers exist to effective government that lawmakers may not know about.

But it’s also about showing Capitol Hill what’s possible, said Anne Meeker, senior advisor for the POPVOX Foundation, which collaborated with the Niskanen Center, Civil Service Strong, the Partnership for Public Service and the Foundation for American Innovation on the work, called Departure Dialogues.

Communication between those making laws and those implementing them has been a problem for a long time, said Meeker, in part because federal employees doing the work aren’t usually authorized to go talk to Congress about what they may be dealing with as they turn statute into reality. That made last year a unique opportunity, as federal employees left the government en masse under the Trump administration's efforts to shrink the size of the federal workforce.

Typically, Congress’ current mechanisms for feedback from those closest to implementation in the government are limited. Hearings “are performative as often as they are informative,” the new report reads. Audits from the Government Accountability Office are usually retrospective, and congressionally-mandated reports are often compliance exercises.

Federal agencies have legislative affairs offices, but the information they transmit to Congress often gets filtered down to politics and top-level priorities, not the “program-level, operational oversights [of] what’s working, what’s breaking, what statutory language creates unnecessary friction,” the new report reads.

The hope is that Congress may replicate POPVOX’s process, or parts of it — including the use of AI to synthesize insights and surface patterns — so that experts have channels to communicate with lawmakers and their staff. For that reason, Departure Dialogues includes a methods report, in addition to a report on key findings and a legislative index.

One recommendation is for congressional committees to consider building structured input processes into reauthorization cycles or invite mid-level experts in for structured listening sessions.

As for what lawmakers may find if they take on the charge of hearing more from those in government agencies, one top takeaway Departure Dialogues found is that the accumulation of policies and requirements is making it difficult to get things done. Congress usually adds requirements, but it doesn’t take them away.

“One recurring challenge I encountered was the cumulative burden imposed by overlapping and sometimes conflicting legislative and reporting requirements associated with different funding streams,” Vikki Stein, who worked at the U.S. Agency for International Development for 30 years, told POPVOX.

“While each requirement was reasonable in itself, together they created inefficiencies that reduced our ability to focus on program effectiveness,” she continued. “This led to significant duplication of effort, diverting staff time and resources away from program monitoring, learning, and adaptation.”

Others told POPVOX that statutory language sometimes directly prevents them from achieving what Congress intended.

A common culprit: the Paperwork Reduction Act, a law created before the days of the internet that’s meant to reduce paperwork for Americans. Detractors say the law adds bureaucracy internally to those delivering government services who want to collect data like feedback meant to help ensure that government programs work well for people — but that it doesn’t ultimately always reduce burden on citizens the way it’s intended to.

Those writing the new report saw themes across workforce, contracting and internal communication: just as agencies and Congress are siloed, so too are agencies isolated from each other, and even teams within agencies are experiencing separation.

“The project was a little bit of an experiment to see if departing federal employees in this really politically tense moment were interested in participating,” Meeker said of the work.

“The response that we got … was really neat to see. We had so many folks really excited about this chance to say to Congress, like, ‘Look, forget the partisanship, forget the politics. This is just the one thing you need to know about how to make this program better,’” she continued. “That spirit of service was actually really kind of moving.”

]]>

Education tech chief heads to OMB as deputy federal CIO

Natalie Alms — Thu, 07 May 2026 12:27:17 -0400

Federal Chief Information Officer Gregory Barbaccia has a new deputy: Thomas Flagg, currently the CIO for the Education Department, according to an internal email viewed by Nextgov/FCW.

Flagg has been at Education since fall 2024, and he worked at the Labor Department for over 11 years before that.

“He understands firsthand the operational realities, constraints, frustrations, and opportunities that agency technology leaders face every day,” Barbaccia wrote in the email announcing Flagg’s new role. “I am looking forward to working with him as he brings that ground-truth knowledge into the federal policy process, helping us better interpret agency pain points and translate them into practical, effective government-wide direction.”

OMB did not return a request for comment.

Flagg joins the OMB team as Barbaccia is split between several roles, including serving as the acting director of the Technology Transformation Services within the General Services Administration. That team helps agencies across the government with their technology, including by maintaining some central solutions. It has lost about 70% of its staff since President Donald Trump took office last year.

Barbaccia is also the federal chief AI officer — overseeing the government’s AI efforts — and serves as the federal government’s service delivery lead under the Government Service Delivery Improvement Act.

]]>

White House weighs reining in contractors’ control over how agencies use AI

Alexandra Kelley and David DiMolfetta — Wed, 06 May 2026 14:11:03 -0400

The federal government is circulating draft policy documents that contain language clarifying the government’s ability to use private sector technology without outside stipulations for how they do so, two sources familiar with their development told Nextgov/FCW.

While it remains unclear if the language being passed between various government agencies — namely the Department of Defense and components of the Trump administration — will manifest into an executive order or finalized policy, that language centers on ensuring the government has control over how its acquired technology products are used.

One source familiar with the ongoing development told Nextgov/FCW that the goal of the language is to clarify that “it is for that democratically elected government to determine what is a lawful and appropriate use of a particular technology, not solely a company.”

The White House is mulling an executive order that would create a working group for AI models before they are deployed, according to a person familiar with the matter. The New York Times first reported the administration’s consideration of the order. It’s not clear if the contracting language is a separate initiative or would be a provision embedded into a forthcoming directive.

Other language featured in the draft documents examines how the government can manage emerging cybersecurity threats posed by AI models like Anthropic’s Mythos Preview and OpenAI’s GPT 5.5, according to the same source and another person familiar with the matter.

The discussions and draft documents highlight how the Trump administration is looking to take a more hands-on approach on the AI sector, despite prior policy positions that signaled a more permissive environment for the evolving technology.

“There is likely going to be another wave of AI government statements,” the first source said.

When asked to confirm the existence of these documents, a White House official told Nextgov/FCW that “any policy announcement will come directly from the President. Discussion about potential executive orders or policy directives are pure speculation.” The Department of Defense referred questions to the White House.

The Trump administration’s efforts to refine the government’s rights when licensing private sector AI models and systems follow a dispute between Anthropic and the Department of Defense over using the company’s AI products in autonomous weaponry and domestic surveillance.

The Pentagon designated Anthropic a supply chain risk, and federal agencies were subsequently required to offload the company’s products from federal workloads. Some lawmakers took issue with the perceived retaliation on behalf of the administration, and Rep. Sam Liccardo, D-Calif., attempted to amend the Defense Production Act to prevent government blacklisting in March.

The administration has telegraphed some of its wants for the relationship between vendors and industries since that debacle, with Emil Michael, the undersecretary of defense for research and engineering, saying on a March episode of the All-In podcast that “all lawful use seems like a good thing” to benchmark against.

Anthropic model capabilities piqued government officials' interest, however, when the company announced the release of its new high-powered Mythos Preview model and associated Project Glasswing for select companies to test in their digital networks.

Leadership at the Pentagon has not been wholly opposed to guardrails on tech use in defense and warfighter operations. Michael told CNBC on May 1 that the Pentagon wants guardrails “in some ways,” but maintained that these guardrails have to align with the government’s needs.

“When they deploy on our networks, they’re deploying models that are tuned for national security purposes,” Michael said. “And that's why the partnership with the executive team and the management is so important, because things are evolving in the threat landscape. And whatever guardrails, whatever principles they want to develop against, has to be consistent with our values, our mandate, our restrictions, even, and that’s where the guardrails come in.”

]]>

Agencies eye agentic AI but readiness questions linger

Frank Konkel — Wed, 06 May 2026 08:00:00 -0400

Agentic AI has entered the chat for federal agencies.

According to a March survey of more than 200 technology executives across government, more than half (53%) said their agencies are exploring agentic AI or actively planning pilots of the technology. Another 15% are currently implementing agentic AI systems or have completely done so already, compared to 6% who said they were not yet considering agentic AI.

However, even as agencies race toward agentic AI, respondents identified potential barriers to adoption in the forms of inadequate oversight policies and disparity between the perceived necessity of governance frameworks and their implementation. The findings were released today by Market Connections on behalf of ServiceNow.

“This research confirms what we're hearing from agencies every day — the appetite for agentic AI is real, but oversight hasn't kept pace," said Mike Hurt, Global Vice President of U.S. Public Sector at ServiceNow. "Seventy-seven percent of federal leaders say oversight frameworks are essential, yet fewer than a third have actually implemented them. Agencies that build accountability into their AI workflows from the start, not as an afterthought, will be the ones delivering strong results for citizens.”

The findings come after a major uptick in AI use across the federal government in 2025 despite a significant decrease in the total number of federal employees.

In April, the Office of Management and Budget unveiled its 2025 Federal Agency Artificial Intelligence Use Case Inventory, which indicated AI use more than doubled across federal agencies from 2024. In total, agencies reported more than 3,000 AI use cases, with significant jumps in AI use at NASA and the departments of Health and Human Services, Veterans Affairs, Justice and Energy.

Agentic AI is generally defined as autonomous systems capable of pursuing complex goals and reasoning, with the ability to take independent actions across software systems with minimal human oversight. Those agentic capabilities to perform some tasks without human intervention have made it an attractive option, and it has been touted by some of President Donald Trump’s top tech officials as a key way to do more with less.

But per the findings, not every agency is ready for agentic AI, even as a growing number of companies offer agentic solutions. Only 20% of respondents said their agencies have defined policies for pre-deployment testing or generic agentic AI use, and only 8% have a defined framework for incident response. Even fewer (6%) have a framework for third-party or vendor governance.

"Agentic AI has definitely entered the chat. Over half of those we surveyed are currently in the planning stages or have actively launched a pilot effort,” said Aaron Heffron, president of Insights and Research at GovExec. “The main question remains, however, if the current infrastructure, both human and technical, is up to the task.”

Findings also indicated some agencies struggle moving agentic AI pilots from the sandbox to production environments. Those challenges move beyond policy and into data readiness.

“One of the very big questions that you have to ask is, ‘How am I getting my data ready for AI consumption? That governance piece becomes critical [to] making sure that your data within your organization and your AI are working together,” advised one unnamed IT director in his response.

ServiceNow Federal Chief Technology Officer Jon Alboum said one way to address the data problem is bringing data and workflows together in a single environment.

“The current environment of fragmented, siloed systems and disconnected workflows only increases complexity and hinders adoption,” Alboum said. “To move forward, AI adoption should focus on bringing everything together in an AI control tower so that policies can be applied, controls enforced, and results delivered efficiently across the organization.”

Research findings indicated that while governance frameworks tended to lack maturity, consistent oversight was a near-universal requirement. Almost 90% of respondents said they required logging and audit trails for all actions, and more than 80% requiring automated policy checks and guardrails.

The findings further point to a strong correlation between the demand for human oversight and criticality of data. For national security, critical infrastructure and emergency response data, 79% of respondents said their agencies mandated “human-in-the-loop” oversight, with approval needed for every action performed by AI. For high-risk data, like benefits claims or agency financial data, 78% of respondents said their agency requires formal human approval before high-risk actions are taken by AI, but not every action. Conversely, more than 90% of those surveyed said they favored reduced direct involvement for low and moderate risk data, requiring only periodic check-ins.

To ensure accountability for agentic AI solutions, 84% of respondents said their agencies had documented escalation policies, while 78% had structured post-incident review processes. Fewer than half (44%) said their agencies included liability or responsibility clauses for AI vendors in contracts, and fewer than one-third (29%) had documented “kill switch” procedures.

“Federal leaders say they want human control over high-risk AI — but less than a third have a kill switch to enforce it, leaving a dangerous gap between intent and capability where trust is won or lost,” the report states. “Agencies must act now to define intervention triggers, ensure data readiness, and unify oversight into a single platform — or risk losing control as systems scale. When failures happen, they cannot be crises; they must be contained, repeatable workflows. Built-in accountability is no longer optional — it is the prerequisite for any agency serious about deploying agentic AI at mission speed.”

Editor’s note: Market Connections is a business division of GovExec, the parent company of Nextgov/FCW.

]]>

Agency leader says AI is helping resource-strained workforce identify more fraud

Sean Michael Newhouse — Tue, 05 May 2026 15:01:28 -0400

An anti-fraud official from the Centers for Medicare and Medicaid Services said that artificial intelligence is enhancing her workforce’s capability to spot scams.

“There's a lot of fraud in the healthcare sector. The estimates on the conservative side are about $100 billion and, depending on who else you talk to, you can easily double or triple that with different calculations,” said Jeneen Iwugo — the acting director for the CMS Center for Program Integrity — at the UiPath Public Sector Summit on Tuesday. “I have a modest budget of $1 billion, so the size of the problem is much bigger than the budget I'm given to find it.”

Specifically, Iwugo explained that AI tools are helping CPI’s roughly 500 employees better identify fraud across the four to five million claims they review every day.

“My team uses AI to comb through those claims and that data and figure out where the risk is the greatest,” she said. “I cannot investigate everything that looks weird, so I have to restratify my work to make sure that I am auditing and reviewing those instances where we have the biggest risk for something fraudulent happening.”

Iwugo emphasized that the Trump administration’s prioritization of combating fraud has given her office more flexibility.

“The longer leash I get, the more I'm able to push the needle with using AI, getting into the agentic AI space, the more of that $100 billion I'll be able to recapture,” she said.

Going forward, Iwugo stated that CMS is piloting programs in which AI doesn’t just review potentially fraudulent claims but also recommends to employees what the possible penalties could be.

“Once I get there, I will be able to take off and capture a lot more of the fraud that I know exists, that I'm able to detect, but that I'm just watching because I can't move fast enough,” she said.

Iwugo also touted that CPI in fiscal 2024 saved Medicare an estimated $26.3 billion, which is a return on investment of $14.6 for every $1 invested and an improvement from fiscal 2023 when the agency projected that it recovered $14.9 billion.

Agencies reported there were 3,611 individual AI use cases in 2025, which is more than double from 2024. The Veterans Affairs Department, as one example, is using the technology to speed up processing of veterans’ benefits claims, but congressional Democrats argued that it is worsening error rates.

]]>

How State finally got online passport renewal to stick

Natalie Alms — Tue, 05 May 2026 11:52:00 -0400

In 2024, the State Department opened its online passport renewal platform, upending the paper-based, 1970s process the department had been intending to revamp for over a decade with little success.

The department has now issued over 7.3 million passports through the online system. Compared against a general discourse about government systems being old, clunky and frustrating, the online renewal tool appears to be a bright spot, as 94% of users have rated it positively in government surveys, according to Matt Pierce, deputy assistant secretary for passport services, consular affairs.

And unlike some of the other recent stories of successful digital government rollouts that did not survive into a new administration, most notably the IRS Direct File program, State’s system is still going well into President Donald Trump’s second term, and the department is planning for more.

State wants to pilot online applications for those seeking their first passport, and it’s looking into issuing digital travel credentials, too, said Pierce, who recently spoke with Nextgov/FCW about what made the online renewal process work and what’s next.

Up until 2024, the process for renewing a passport was largely the same as it had been for decades, even as the number of people getting passports has been increasing. In 1990, only 5% of Americans had a passport. Now, that number sits at around 50% — and that figure is expected to continue to grow, said Pierce.

State debuted its first attempt at an online system in 2022, in the lead-up to record passport backlogs the following year, as Americans looked to travel again following the COVID pandemic. At the time, the department was still grappling with staffing shortages caused by a hiring freeze instituted during Trump’s first term.

That pilot worked for some people but was ultimately unsuccessful, one former State Department employee who worked on online passport renewals told Nextgov/FCW. They requested anonymity for fear of retribution.

In technical terms, the first system was made in a waterfall development style. State made a list of requirements and chucked them “over the fence” to technologists who built a tool, they said.

Passport adjudicators weren’t consulted in the design process and had a really difficult time using the system. Applications would get lost because of how work queues were set up, the former employee said.

The department paused, pivoted and eventually opened the system that’s still running now in 2024.

What changed? For one, the department switched to a human-centered, agile design process.

There are thousands of employees who handle over 24 million passports a year, Pierce said, adding “that system has to work for them.”

In the lead-up to the second launch, the team tested the system with frontline employees and worked to make sure that the understanding of those building the new system matched up with the needs of those actually handling the day-to-day passport work. That included considering not only technology alone, but also processes and policies.

And while the department attempted to replace the entire system in the first rollout, for the second, they replaced only the front end portion that Americans see. On the backend, enhancements were made, but the system is largely as it was before, meaning employees didn’t have to make changes, said Pierce, while the department continues to work on a larger overhaul.

Moving forward, Pierce is looking to use the same dialogue with employees that made the second online passport renewal effort successful to find improvements to prevent future backlogs along with a staffing baseline. It’s an organizational transformation, he said, driven by employees to change processes and procedures.

One big shift for the department, he said, has actually been a cultural one in how it responds to risk.

“In the past, we would spend a lot of time thinking about something, working on something, and then here it is. Now you’ve got to live with it for 10 years,” Pierce explained.

Now, employees are more open to taking calculated risks with the understanding that they can pilot changes and adapt as needed, he said, noting “that has been a huge change.”

More improvements ahead

Over a year out from the second, more successful launch, it takes 20 minutes to renew a passport on the new, online system, as opposed to the 40 minutes it took through the old process, said Pierce. The department estimates that it’s saved Americans over a million hours.

The online system currently handles over half of renewals. Some people aren’t able to apply online depending on their situation, although the goal is to eventually make it so that anyone renewing can do so online, said Pierce.

The team is also planning to pilot letting Americans that want to get their first passport apply online in the coming years. That’ll require the department to work through some wonky issues, like how it will digitally validate proof of citizenship documents that State itself doesn’t house, like birth certificates — something that will likely require the department to work out data-sharing agreements with states, said Pierce.

State is also in the early stages of looking into digital travel credentials.

“A lot of people call it the digital passport,” said Pierce, but it’s different from the ID that digital wallet users can create with a passport, for example, which can’t be used for international travel or border crossings.

“I mean something that can ping against the database, like the passport, to validate that it is a valid passport issued by the United States government, and this is the person’s information,” he said.

The department is also working to add Trump’s face to a limited number of commemorative passports, although that news broke after Nextgov/FCW’s interview with Pierce.

New challenges

As it pursues new changes, the department will be working with a different team than the one it used for online passport renewal.

Last year, Pierce won what’s considered the Oscar of government service, a Sammie, alongside another key leader in the passport modernization effort, Luis Coronado, the former CIO for the Bureau of Consular Affairs at State. But Coronado and others have since left the department.

The Bureau of Consular Affairs wasn’t spared from layoffs last year as the Trump administration sought to downsize the federal workforce, although some affected employees were later reinstated.

Other offices that worked on the online passport renewal project have also been shuffled around as part of a reorganization. Consular Affairs’ tech office, led by Coronado, was moved to the central IT department within State, and also was affected by layoffs, said the former employee, who has also left federal service.

“The intent of the reorg is to become a more responsive, a more relevant State Department,” said Pierce, adding that he’s seen “a lot of positives.”

Other departments, including Interior, have also been centralizing their technology operations.

The team behind the online passport renewal launch also included 18F, a digital services consultancy inside the government that the Trump administration shuttered completely last year, as well as the U.S. Digital Service, which was renamed to house the Department of Government Efficiency as the U.S. DOGE Service.

“We lost a lot of the folks that had been integral to getting that online passport renewal successfully stood up and out the door for the 2.0 version,” said the former employee.

These teams helped other agencies adopt different models of modernizing, such as the IRS with Direct File.

“The federal government has lost a lot of that,” the former employee said. “How is the government — absent those organizations — going to continue to transform how the government improves services?”

]]>

Growing agency talent is critical for modernization, Transportation’s IT head says

Edward Graham — Tue, 05 May 2026 08:00:00 -0400

Building out new technology capabilities within the Transportation Department is the relatively easy part of modernization; working to ensure that personnel have the resources and training they need to leverage the tools effectively is one of the real challenges, the agency’s digital and IT lead said on Monday.

During a keynote speech at GITEC 2026, Transportation’s chief digital and information officer, Pavan Pidugu, said the agency’s “1DOT” initiative has been focused on enhancing the efficiency of its IT infrastructure. Pidugu was a 2026 Fed100 winner for driving that initiative by unifying disparate internal organizations and systems into the streamlined 1DOT operating model.

“We're simplifying the way we work,” Pidugu said about the effort. “We're simplifying the assets that we have, the technology applications we have, and we are unifying our processes to make the best use of technology that's available today.”

While this includes streamlining processes through modernization — not an easy lift, by any means — Pidugu said “building the new technology probably is the easiest thing” when it comes to changing business-as-usual across Transportation.

“What are the challenges that we face today? While we are in this major transformational journey, there are lots of gaps, and our people, our processes need to change,” he said.

To not frustrate agency employees during this modernization effort, Pidugu said Transportation has been looking to use the process to build out opportunities for employees.

"We're going through a monumental and historic change that the department has never encountered or went through,” he added. “How can we make it fun? How can we help our people achieve their personal and professional goals?”

To accentuate this process, Pidugu said the agency is “creating lots of new opportunities for our workforce in growing our talent.”

Transportation’s most recent IT strategic plan, which delves into the 1DOT initiative, includes a focus on growing agency talent alongside the broader push to eliminate its technical debt.

The document says the agency is looking to accomplish this goal by pushing to “invest in our Federal workforce’s technical capability via skills assessments, targeted training (cloud-native, Agile, data/AI, security), rotations and mentorship, competency-based hiring, and performance-based recognition to retain and scale institutional expertise.”

To further empower its workforce and grow talent, Pidugu said Transportation has also been looking to industry to help with some of these efforts.

“Not everything can be done within the DOT in-house,” he added. “We're a small organization. We have limited resources as appropriated through the funding channels. There is a lot of work to be done, so we are trying to make sure, ‘where do we establish those partnerships with the industry?’ We're making sure what innovation we can influence in the industry to support our needs. And through those partnerships, we're just beginning.”

But, right now, Pidugu said internal buy-in from the entire department is the major key to success.

“There should be a lot of support from leadership all the way from the top,” he said. “There should be a lot of excitement all the way from the bottom also. When there is support and excitement, then you can do magic.”

]]>

One CEO’s push to speed up how the Pentagon adopts technology

Heather Kuldell-Ware — Fri, 01 May 2026 13:59:39 -0400

Early in her career, Meagan Metzger walked into a program office with a simple question: Why was the government building software it could just buy?

Again and again, she saw the same pattern: the government spending time and money building technology that already existed in the commercial market — often ending up with something slower, more expensive and less effective.

She saw it as a systems integrator building software for the Army, scaling a company and working with commercial tech firms that wouldn’t touch the federal market. And she saw it inside government, where the challenge wasn’t the technology.

“The tech’s easy,” she said. “It’s all the unsexy stuff that gets in the way.”

So she started figuring out how to work around it. Where others saw barriers — requirements, contracting, incentives — Metzger saw problems to solve. If there was a path forward, she was going to find it.

In 2015, she founded Dcode, a tech accelerator, to help startups break into the federal market. But over time, she realized the bigger problem wasn’t companies, it was the system. Today, Dcode works with government leaders to help them become better customers of technology.

A decade later, that work had taken on new urgency. Metzger worked directly with leaders across the U.S. Army and the Defense Department to turn acquisition reform into execution, helping them act on new authorities as they took shape.

As reform accelerated, Metzger pushed leaders to think differently: not program by program, but as a portfolio. What matters most? Where do you take risk? What do you stop doing?

Instead of waiting for perfect guidance, she helped teams build ways to act through decision structures, operating rhythms and mechanisms that let them move as soon as the door opened.

The results were fast. Dcode helped drive an OTA-based procurement from zero to live in under 30 days, launched a secure tech marketplace proof of concept in 24 hours and connected dozens of venture firms and hundreds of companies to urgent mission needs.

More importantly, Metzger helped close the gap between identifying commercial technology and getting it into operators' hands by cutting through the friction that typically slows it down.

At times, that work has put her in rooms where few outsiders are invited. In one meeting with more than 20 one-, two- and three-star generals, Metzger was the only civilian — and one of the only women — at the table.

At one point, a senior leader paused the discussion and asked what they were doing wrong.

“I was just like, ‘China does not care about the decisions you're making right now. They're going to kick your ass,’” she said. “They trusted me to be in that room and call them on the carpet, but then want to do something about it.”

For Metzger, that’s the job: find the friction and remove it.

]]>

GSA procurement chief leads FAR overhaul

Heather Kuldell-Ware — Fri, 01 May 2026 13:58:51 -0400

Jeffrey Koses paced the floor of his home in the middle of the night in 1995, trying to get his newborn son back to sleep by talking him through potential responses to a stack of procurement protests waiting on his desk.

Somewhere in those 2 a.m. conversations, three ideas took hold. The government needed to communicate better with industry to avoid protests. The workforce shouldn’t be stuck working 16-hour days just to keep up. And technology should make the job easier, not harder.

Today, as the senior procurement executive at the General Services Administration, Koses applies those same principles to one of the government’s most ambitious acquisition reforms: the Revolutionary Federal Acquisition Regulation Overhaul.

Koses’ path into federal acquisition wasn’t planned. With a degree in history and political science and “no clue” what to do next, he landed in a GSA management development program buying furniture for the federal government.

“I found the mission was just fascinating, because GSA is the behind-the-scenes engine that keeps the rest of government working,” he said. “Any initiative that I saw going on in the country, any story in the paper, I would find a GSA connection.”

He moved between operations and writing policy, often having to live with what he wrote. Policy, he learned, isn’t something people sit down to read; it has to be clearly written for the contracting officers and industry partners who use it.

That personal experience shaped his approach to major initiatives. Koses served as the executive sponsor behind OASIS, a professional services contract vehicle that first launched the scorecard model and became one of GSA’s flagship programs. He also introduced transactional data reporting to the Federal Supply Schedules, shifting pricing toward real-world marketplace data.

Koses recognized early that the real challenge in rewriting the FAR would be driving a culture change.

Instead of relying solely on the traditional rulemaking process, he built a cross-agency “tiger team,” brought practitioners and technologists into the process, and used deviations to test changes in real time.

Within six months, the text of the FAR was reduced by 25%, eliminating 484 pages and 2,724 “must-do” requirements and giving acquisition professionals greater flexibility to make judgment-based decisions.

That flexibility, Koses believes, will change how the system operates. He paired the overhaul with a push for transparency and engagement, such as industry-only sessions designed to encourage candid feedback on public platforms like Acquisition.gov.

For Koses, that work reflects a broader leadership approach: go where the problems are.

“I always looked for the office or the program that was seen as troubled,” he said. “If I go to something that’s struggling, I can make it better.”

Decades after that sleepless night, the throughline remains the same: simplify the system, empower the people using it and make government work better from the inside out.

]]>

Federal workforce trauma is creating a stumbling block for AI adoption

Howard Risher — Thu, 30 Apr 2026 07:00:00 -0400

The federal government and the U.S. economy is at a crossroad. There are two contrary but seemingly independent developments that have profound implications for the workforce. Both sides cannot be correct but to build support for what’s unfolding the differences need to be understood and resolved.

One is the rapid transformation of jobs and work systems driven by AI. The frequent headlines contend the drive to roll out AI is the future. It’s inevitable. The titles suggest a solidly optimistic view that AI will be:

“a Net Economic Accelerator,”
“a Force for Better Work, Not Job Loss,”
“a Force for Better Work, Not Job Loss,”
“a Driver of Safer, More Efficient Daily Life.”

The contrary argument has made the headlines of the country’s more prominent websites:

CNN — “Tech industry experts warn AI will make us worse ...”
Forbes — “The Real Threat of AI: WEF Global Risks Report 2025”
BBC — “Artificial intelligence could lead to extinction, experts ...”
The New York Times — “The A.I. Prompt That Could End the World”

There have also been reports and columns arguing AI has revealed problems in how government work is organized and managed, and from a GAO report, that the work models are not suited to AI integration.

DOGE made the problems worse

Elon Musk has been very vocal. On one side, he contends AI will create unprecedented abundance, producing goods and services “far in excess” of current levels. He has also argued AI will eliminate most or all human jobs. When that happens, he promoted a form of government-provide “universal high income.”

Musk is relevant of course because of what he and the Department of Government Efficiency (DOGE) did to the federal workforce. Through 2025 DOGE initiated (1) mandatory attrition targets, (2) reductions-in-force plans, (3) a hiring freeze and (4) contract cancellations that removed contractor employees. The actions triggered 348,219 individuals to quit, retire, were laid off or otherwise left federal employment. At the same time, 116,912 people started working for the federal government — a 55.6% decrease from the year before. The net reduction, according to the Pew Research Center, was nearly 238,000.

The impact on the workforce has been pronounced. There has been a loss of institutional knowledge, support functions disappeared, worker shortages have impeded agency performance, and reports contend agencies are struggling to maintain mission delivery. The Brookings Institute expert, Elaine Kamarck, reported the cuts made agencies “scramble to fill critical gaps in services” ... There are time bombs all over the place ... They’ve wreaked havoc across nearly every agency.”

Employees are not open to AI

When the Office of Personnel Management canceled the Federal Employee Viewpoint Survey for 2025, the Partnership for the Public Service created and conducted a similar survey late in the year. It had responses from 11,083 employees across executive branch agencies.

The survey “revealed significant challenges” — read significant declines — in federal employee engagement and morale in 2025. The Partnership’s CEO, Max Stier, made it clear what the decline means:

“We have every red light blinking across the federal government. Morale is as low as imaginable.”

“This workforce has been fundamentally traumatized ... That’s not good for anyone. It’s bad for the workforce, it’s fundamentally bad for the American people, and it will lead to use be less safe, healthy, and prosperous as a society.”

“This loss of expertise directly harms Americans’ access to critical services and will take decades to repair. [The losses leave] dangerous gaps in key federal services, like food safety inspection, Social Security processing, veterans’ health care and disaster response.”

Research in other sectors by Gallup and others shows clearly a demoralized workforce triggers a high cost. In Gallup’s terms, that is when employees are “actively disengaged” or “unhappy and unproductive at work.” There is no direct comparison but the Partnership’s survey shows the “chainsaw” workforce cuts left the workforce “traumatized.” Psychological safety “collapsed.” It could hardly be worse.

The Partnerships analyses show the “reforms to the federal government cost the U.S. economy more than $165.6 billion ...” Possibly more important going forward is the loss of the better performance when employees are fully engaged. Gallup has promoted the value of engaged employees for over three decades. Their research has linked engagement levels to a long list of employee performance metrics.

Gallup “research has repeatedly shown that engaged employees are the lifeblood of successful organizations. They are not just loyal and productive; they the driving force behind innovation and customer satisfaction.” Their research “... reveals a stark contrast between teams with highly engaged employees and those struggling with disengagement.” Companies with an engaged workforce are more productive, more profitable and have higher customer satisfaction.

OPM has reported its measure of employee engagement for years but for unclear reasons has never reported finding a connection between employee engagement and performance. However, it is very clear that is no longer a consideration.

Research also shows clearly disengaged workers are not open and supportive of implementing AI. That point was emphasized in a recent "Forbes" column, “Why You Can’t Lead an AI Revolution Without Engaged Managers.” A column summarizing Glassdoor reviews and social media posts related to AI concluded many “employees are pretending to use AI tools just to comply with internal protocols ... multiple people admitted to exaggerating or fabricating their usage.”

Projected national job losses

From a broader perspective, researchers at Tuft’s Fletcher School of Global Affairs recently released the American AI Jobs Risk Index. It summarized the projected job losses for 784 occupations in 20 industries.

“It is a first-of-its-kind data-driven framework that maps the potential of AI-driven job vulnerability across every major occupation, industry, metropolitan area and state in the United States. ... the Index goes beyond prior studies by measuring actual vulnerability to job loss — not merely exposure — and connecting that vulnerability directly to projected income loss and geography.”

“The Index projects approximately 9.3 million U.S. jobs are at risk of displacement in the next 2–5 years, with a plausible range of 2.7 to 19.5 million depending on alternative adoption scenarios. Associated household income at risk spans $200 billion to $1.5 trillion annually ... equivalent to the economies of Belgium and, under faster AI adoption, approaching that of South Korea.”

“Industry-wide vulnerability averages approximately 6%, but the steepest risks sit in Information (18%), Finance and Insurance (16%) and Professional, Scientific, and Technical Services (16%).”

Based on the Index estimates, the country will lose from the three groups a total of over 3 million jobs. The federal government has an AI estimated 128,500 Information specialists (in 14 job series) and would lose 23,000 specialists.

The biggest losses are projected to be in California, Texas, New York, Florida and Illinois. The losses are also high in the District of Columbia. “Wired Belts” like Ann Arbor and Boulder could become “Rust Belts.” That will have political consequences.

The economic concerns include:

Job displacement and lost income
Wage suppression prompted by reduced labor demand
Increased economic inequality benefiting capital investors
Erosion of human skills and loss of adaptability

Microsoft co-founder Bill Gates’ made a disturbing AI prophecy: “Humans won’t be needed ‘for most things' in 10 years.”

Public policy debate: Regulation needed?

As long ago as 2021, the European Commission, the executive arm of the European Union, proposed the first EU artificial intelligence law, establishing a risk-based AI classification system. It did not pass initially but in 2024 the EU Artificial Intelligence Act became the world’s first comprehensive AI law.

In contrast, President Trump released a promised “AI Action Plan” in July 2025 that outlined “over 90 federal actions focused on three areas of focus: increasing private-sector innovation, expanding AI-related infrastructure and exporting American AI.” He followed that with three Executive Orders promoting American AI products.

In December he signed an EO intended to override certain state laws on AI, with department heads to identify the laws deemed “burdensome.” All 50 states considered AI-related measures in 2025.

A number of prominent scholars and economists have argued legislation is needed. A core issue is — developments in AI are ongoing and it’s not possible to anticipate in advance what may be warranted. Tomorrow’s AI systems could be very different. It is clear that the legal process is slow, incentives are misaligned and the complexity of AI systems is growing faster than governance.

Questions for Elon Musk: When will the economy warrant creating the government-provided universal high income? Does it ride on high unemployment or low family income? How will it be funded?

]]>

CBP seeks AI solutions to keep pace with rising volumes of border scans

Nick Wakeman — Wed, 29 Apr 2026 17:46:47 -0400

Customs and Border Protection is facing a volume problem as it expands the scanning of cars and trucks at ports of entry.

The so-called non-intrusive inspection scanning technology that CBP is deploying greatly increases the number of vehicles scanned.

“The sheer volume of data generated by increased scanning can be overwhelming for human operators, necessitating innovative solutions to maintain efficiency and effectiveness,” the agency writes in a sources sought notice posted Wednesday.

CBP uses X-rays to scan full-size vehicles – commercial trucks, cargo containers, and privately owned cars – without physically opening or unloading them. The agency's systems generate tens of thousands of images a day across multiple types of scanners.

To address this volume challenge, CBP wants to use more artificial intelligence and machine learning solutions. But the agency does not want a proprietary data platform, data pipeline or case management system.

Instead, the agency wants algorithmic components that can be integrated with existing and planned CBP-owned data platforms.

CBP is looking for help with using a human-in-the-loop approach. AI will serve as a tool as “CBP’s frontline personnel remain the most critical asset to thwarting illicit goods crossing our borders,” the request for information states.

While the agency emphasizes in the notice that employees are its most valued resource, “CBP recognizes that technology is a vital element to mission success.”

The agency wants information on algorithm products, including performance metrics related to:

Anomaly detection
Manifest/commodity verification
Identifying modifications, such as concealment locations
Contraband detection.

Responses are due May 30.

]]>

White House is drafting plans to permit federal Anthropic use

Alexandra Kelley and David DiMolfetta — Wed, 29 Apr 2026 13:39:06 -0400

The White House is crafting guidance that would allow federal agencies to bypass a supply chain risk designation on Anthropic and clear the way for government use of its tools, including the cyber-focused Mythos AI model, according to an industry source familiar with the matter.

The move is notable because it suggests the Trump administration may be softening its stance on Anthropic. The Pentagon labeled Anthropic as a supply chain risk earlier this year — and the White House later ordered a governmentwide phaseout of its technology — after the AI company declined to ease restrictions on its products being used in domestic surveillance and fully autonomous weapons.

The administration is also drafting an AI executive order that could, in part, address how the government uses Anthropic's tools, a second person familiar with the matter said.

Both sources spoke on the condition of anonymity to communicate non-public details. Nextgov/FCW has asked Anthropic for comment.

"The White House continues to proactively engage across government and industry to protect our country and the American people, including by working with frontier AI labs," a White House official told Nextgov/FCW. "The collective effort of all involved will ultimately benefit our economy and country. However, any policy announcement will come directly from the President and anything else is pure speculation."

Axios first reported on the White House's plans.

Mythos, unveiled earlier this year, has become a tipping point for cybersecurity and AI practitioners because it demonstrates how advanced models can be purpose-built for real-world cyber operations, including those planned inside the intelligence community. In the wrong hands, it could be used to carry out sophisticated cyberattacks against government networks, critical infrastructure or other key U.S. systems.

As soon as Mythos was unveiled, the company launched Project Glasswing, a private coalition of firms that aims to use the model to patch critical vulnerabilities across the global internet before AI-assisted cyber threats become widespread.

Last week, Joint Chiefs Chairman Dan Caine said at a national security event in Nashville that autonomous weapons will be a “key and essential part of everything we do,” pointing to a broader military push to integrate AI into defense operations.

“I want our nation using all of our best [AI] models,” retired Gen. Paul Nakasone, who led NSA and Cyber Command, told reporters in a briefing at that same event hosted by Vanderbilt University.

“I don’t think it was accurate that Anthropic is a supply chain risk,” added Nakasone, who also serves on the board of OpenAI. “I feel uncomfortable with the fact that part of our nation’s capability is not being used by our government.”

The dispute between Anthropic and the Trump administration has rattled Washington’s AI vendor landscape over the last couple of months, with companies scrambling for clarity on contracting requirements as uncertainty grows over how the government will handle how much use of Anthropic products is permissible.

The company has legally challenged the supply chain risk label. A federal judge issued a temporary injunction on the designation and ban in late March, which the government has said it intends to appeal.

Last week, President Donald Trump said in a CNBC interview that the company is “shaping up” and can “be of great use” in the future, a sign that tensions between Anthropic and the government may be easing.

Editor's note: This article has been updated to include comment from the White House.

]]>

FDA to pilot real-time clinical drug trials through cloud and AI

Frank Konkel — Wed, 29 Apr 2026 12:15:17 -0400

Through a new pilot program announced this week, the Food and Drug Administration will use artificial intelligence and cloud computing to monitor clinical trial data in real time, an effort that could ultimately shave years off the approval timelines for new drugs, devices and medications.

FDA commissioner Marty Makary announced “the first ever real-time clinical trial” Tuesday at a press conference held at FDA’s headquarters in Silver Spring, Md. “Today is a milestone day for us to challenge the assumption that it takes 10 to 12 years for a new drug to come to market.”

Makary said about 45% of the time between when a company conducts a Phase 1 clinical trial and submits its applications—which can sometimes be millions of pages long—to the FDA is “dead time,” where investigators and staff are doing “paperwork and other tasks, many of which are tedious.”

The pilot allows the agency to have “a direct data feed from a clinical trial, where the FDA will see what is happening, in the cloud, with the predefined clinical endpoints and any other signals investigators and regulators decide are valuable,” Makary said. “When a patient develops a fever, or a tumor shrinks, FDA regulators can see in the cloud, in real-time, exactly what is happening.”

FDA Chief Artificial Intelligence Officer Jeremy Walsh, who Makary credited as a driving force behind the pilot, said the idea for real-time drug trials manifested last summer through a confluence of the right personnel, emerging technologies and leadership’s drive to modernize the review process, which had not changed much since the 1960s.

Walsh told reporters that “while there is an opportunity to shave off” as much as 40% of the clinical trial time, the agency won’t be cutting corners on safety.

“The goal here is to sort of get to a regulatory decision in a faster timeline, without compromising any safety,” Walsh said. “The goal here is to raise the bar for what can be done. We are reimagining what information we need and when we need it in order to make a decision.”

So far, two clinical trials conducted by pharma companies AstraZeneca and Amgen will pilot the new system.

“The ultimate goal is to move data as quickly as possible across this ecosystem to accelerate our ability to bring new therapies to patients to make a meaningful difference in their lives,” said Amy McKee, oncology senior vice president at AstraZeneca.

In conjunction with the announcement, the FDA released a request for information for the public and industry to solicit input regarding how AI-enabled technologies “can improve efficiency, speed and quality of decision-making in early phase clinical trials.” Responses are due May 29 and could shape an expansion of the pilot this summer.

FDA’s broader tech modernization effort

During the press conference, Makary highlighted several major modernization efforts undertaken by FDA’s IT and AI team, a brain-trust that includes Walsh, acting chief information officer Sridhar Mantha, acting deputy CIO Sanjay Sahoo and acting associate director of business operations Gregory Jackson.

“Thanks to the moderation efforts by our entire IT and AI team here at FDA, we have been able to get away from the fiefdom culture where every center has to have their own license agreement for the same software and their own system,” Makary said. “We have done a massive consolidation.”

Over the past year, FDA consolidated 40 separate application intake systems into a single system, further consolidated its three data monitoring systems and seven adverse event reporting systems each to single systems, and systematically reduced the duplication of various software licenses across the agency’s multiple centers.These consolidation efforts—performed without additional staff or resources after significant staff downsizing in early 2025—“is going to save us at least $120 million a year,” said Makary, noting that money would be reinvested in the scientific community, new technologies and rehiring as many as 3,000 new scientists.

“That added reserve in our resources will support our bold reform agenda,” Makary said.

The agency has also taken massive steps in adopting generative AI.

Walsh, speaking last week at Google Cloud Next in Las Vegas, said in early 2025, about 1% of the agency’s workforce regularly used generative AI in their jobs. Today, the agency has a generative AI adoption rate of more than 80%, with some individual centers exceeding 90%. One of its most popular tools is Elsa, a large language model-powered tool that assists employees with reading, writing and summarizing reports.

FDA makes use of both Google’s Gemini and Anthropic’s Claude models.

“We have deployed enterprise AI across the entire agency, giving our highly skilled, highly-educated workforce access to all these tools,” said Walsh, adding that AI is helping FDA achieve its mission faster. “One of the great things about working with companies like Google in this is the speed at which we can get access to the models where our data sits. When new models are released, we can get access to the model in a week. For us, we need to make sure we have access to the latest capabilities and tools.”

Walsh said in early pilots, AI has already dramatically reduced timelines for data- and documentation-intensive regulatory duties FDA regularly performs, in some cases reducing administrative tasks from 10 days down to 20 minutes.

]]>

Treasury missed security controls in giving DOGE system access, GAO finds

Natalie Alms — Tue, 28 Apr 2026 17:41:16 -0400

Congress’ watchdog reported on Tuesday that the Treasury Department gave a Department of Government Efficiency associate access to the government’s payment systems last year without fully following all of its own security controls — and the DOGE team didn’t always hew to Treasury’s protocols, either.

The findings are among the first reports the Government Accountability Office has released about DOGE’s work, and GAO is working on more audits focused on DOGE access to government systems, a spokesperson confirmed with Nextgov/FCW on Tuesday.

The new report on Treasury zeroes in on atypical access to the government’s payment systems given to DOGE associates.

DOGE access to sensitive government data and systems across agencies has been a flashpoint since the early days of Trump’s second term. DOGE associates said in court testimony earlier this year that pushing for high-level access to government systems was “operating procedure” for the group.

At Treasury, soon after Trump took office last year, individuals on billionaire Elon Musk’s team reportedly began pressing for officials to hand over system access to DOGE employee Tom Krause so that the department could freeze foreign aid payments. A top Treasury official was eventually pushed out of his job after refusing to provide access to the systems.

GAO found that Treasury handed over access to view, copy and print data from the Bureau of Fiscal Service’s three payment systems to an unnamed DOGE associate, who could also see the systems’ source code.

But that member of Musk’s team never completed required security training while working at the department or signed Treasury’s “rules of behavior” policy for IT security while working there.

GAO doesn’t name this DOGE employee in its report, but other details provided by the watchdog match with public reporting on Marko Elez, like the day he resigned, Feb. 6, 2025, following reporting about his racist social media posts. He later went on to work for DOGE at other agencies.

At one point, Treasury accidentally briefly gave that same DOGE employee the ability to make changes in one of those systems, something GAO said was due in part to the agency’s lax procedures and the fact that the access being requested was changed several times before it was approved. The DOGE employee didn’t use the system during this time, GAO says. According to reporting on court records, this was also Elez.

GAO also found that Treasury’s data loss prevention tools didn’t track or block Elez from improperly sending unencrypted information on foreign aid to two DOGE associates at the General Services Administration. Elez did this without getting agency approval for sharing the information on U.S. Agency for International Development payments.

Treasury didn’t discover that incident until it conducted a forensic review of the laptop after Elez had left the department. The department didn’t find the incident sooner in part because its tools aren’t set up to look for information being sent to other government agencies, the report says.

GAO included several recommendations regarding the department’s IT security processes in the audit, only some of which Treasury formally agreed with.

The top Democrat on the Ways and Means Committee, Rep. Richard Neal, D-Mass., said in a statement that “GAO has confirmed our worst fears,” and called on the Treasury to implement all of GAO’s recommendations.

Among those the department didn’t formally agree or disagree with is one urging it to conduct exit interviews and get signatures on post-employment documentation from those with access to sensitive payment systems who leave the department without doing so — including the DOGE employee discussed in the report with access to systems.

The watchdog will be issuing additional reports on DOGE access to Treasury payment systems, it says in the report. The topic has also been working its way through the courts. A district court judge granted a preliminary injunction limiting DOGE access to Treasury systems with sensitive information last year, although that was later modified to allow some access to systems.

GAO also released another report Tuesday on DOGE’s access to systems at the NLRB. Just over a year ago, a whistleblower in the agency said that DOGE had extracted troves of data from the agency using secretive methods during March 2025.

The NLRB’s inspector general has an ongoing investigation into the whistleblower’s declaration, the office confirmed to Nextgov/FCW.

GAO’s new report, however, focuses only on the period after DOGE was detailed into the agency in mid-April, so “to not overlap with the NLRB Inspector General’s Investigation.”

Whistleblower Aid, which is representing the NLRB whistleblower, noted the significance of GAO beginning its review period after their client’s disclosed events took place.

“Because the GAO did not investigate any matters that fell within the timeframe disclosed by our client — in fact scoping it out of their investigation — the report cannot address our client's detailed accounts,” Whistleblower Aid told Nextgov/FCW. “Accordingly, the timeframe investigated by the GAO has no relationship to the wrongdoing witnessed by the whistleblower in February to early April 2025.”

GAO found that the DOGE team asked for access to NLRB systems, but didn’t use the access it was granted. DOGE didn’t even pick up NLRB laptops before their detail agreements expired in July, according to GAO.

]]>

NIST is giving fingerprint examiners better tools for a messy job

John Breeden II — Fri, 24 Apr 2026 14:49:32 -0400

Americans have spent generations watching detectives in dark trenchcoats pore over complex crime scenes in movies and on television. They examine the room, snap photos and break out the familiar blue powder to dust for fingerprints. The ritual is so familiar that it can seem almost automatic.

What those scenes rarely capture is how much effort goes into making fingerprint examination more accurate, more consistent and easier to teach. That quieter work is exactly what the National Institute of Standards and Technology is trying to strengthen.

NIST recently released two resources aimed at helping forensic fingerprint examiners do their jobs better. One is a fully annotated version of NIST’s Special Database 302, a collection of roughly 10,000 latent fingerprint images. The other is what NIST calls OpenLQM, newly created open-source software that helps assess the quality of latent fingerprints and sort them according to how much useful detail they contain. NIST says the two releases are meant to improve forensic fingerprint examination, which remains an important part of many criminal investigations.

Fingerprint analysis is one of those forensic tools that many people assume was perfected long ago. In reality, examiners often work with partial, smudged or otherwise imperfect prints recovered from real-world objects. Training people to evaluate those prints well takes experience, repetition and good examples. It also increasingly requires better ways to train software systems that can assist human examiners without replacing them.

NIST says the newly completed dataset will help train both human examiners and machine learning algorithms to distinguish important features and weigh their value as evidence.

The most vivid part of the NIST fingerprint accuracy project is how ordinary the source material really was. As NIST computer scientist Greg Fiumara explained, “The prints are from people we recruited to come in and do things like write a note, pick up a circuit board, handle a dollar bill, that sort of thing. Then we recovered the prints they left behind using different methods that crime scene investigators commonly use.”

So the new collection is not made up of idealized prints from a textbook. They are the kinds of latent impressions that people leave behind all the time while moving through everyday life.

That realism has been part of the project from the beginning. When NIST first released SD 302 in 2019, it described the database as a set of latent fingerprints left on everyday items by a few hundred volunteers in a lab setting, with other personal information stripped away. The point was not to create a neat archive of perfect examples, but to give researchers and examiners a more realistic way to measure accuracy and test methods against the kinds of prints they actually encounter.

What is new now is that the entire collection has been annotated. Those annotations mark details about fingerprint quality, including regions where ridge patterns are clear, smudged or incomplete.

NIST says those markings make the dataset much more valuable as a teaching tool because they show both humans and algorithms what to look for and what to avoid when evaluating a print. The annotations add structure and interpretive guidance to a dataset that already had broad global use. NIST says more than 1,000 research organizations in more than 90 countries have downloaded the collection since its initial release.

The second part of the project is just as practical. OpenLQM gives examiners a way to score the quality of a latent print on a scale from zero to 100. And it can run as a standard executable or be embedded inside another program or application for maximum portability. The new software can help investigators sort through large volumes of prints and focus their attention first on the ones most likely to contain useful identifying details.

As Fiumara put it, “you give OpenLQM a fingerprint and it returns a number from zero to 100 that is an assessment of the print’s quality.” NIST says the software was adapted from a tool once limited to U.S. law enforcement. It is now being made openly available in a form that can run on Mac, Windows or Linux.

Since the software is open-source and available for anyone to download, NIST is not just improving a government tool for internal use; it’s pushing better forensic resources into the wider scientific and practitioner community. The agency’s biometrics resources page now lists both Special Database 302 and OpenLQM among its available forensic databases and software tools, reinforcing the point that this is part of a broader effort to build reproducible, shareable infrastructure around forensic biometrics.

What makes the fingerprint accuracy project especially useful is that it focuses on the less glamorous side of forensic work. Instead of chasing some dramatic new breakthrough, NIST is improving the underlying tools that fingerprint examiners rely on every day. Better data, clearer annotations and a faster way to assess print quality may not look dramatic from the outside, but they can make difficult work more consistent and efficient.

That is what gives this release its value. It strengthens one of forensic science’s oldest disciplines without pretending to reinvent it. Human judgment still matters, and fingerprint work will probably always involve a measure of skill and interpretation.

So all those movies and TV shows with investigators (still wearing stylish black trenchcoats) dusting for prints will still be accurate — at least for now. But with better training material and advanced tools, that work can become more consistent and easier to teach while also producing more trustworthy results.

John Breeden II is an award-winning journalist and reviewer with over 20 years of experience covering technology. He is the CEO of the Tech Writers Bureau, a group that creates technological thought leadership content for organizations of all sizes. Twitter: @LabGuys

]]>

IRS lacks transparent plans to leverage tech in the face of staffing cuts, GAO and employees say

Natalie Alms — Thu, 23 Apr 2026 17:00:00 -0400

The IRS is banking on using technology to do more with fewer employees. But staff inside the IRS say that how the agency will do that — considering that its own IT shop has lost personnel — is still unclear, and Congress’ watchdog says that the IRS still isn’t being transparent about its long-term tech plans.

The IRS continues to rely on some systems that date back to the 1960s. It’s been trying to modernize them for decades, and was using some of the money from the Inflation Reduction Act to do so under the Biden administration. Congress has since clawed back most of that funding, and the remainder is set to run out in fiscal year 2028.

A few months after President Donald Trump returned to the Oval Office, the IRS paused its modernization work to re-evaluate its strategy. IRS leadership said they wanted to rely more on generative artificial intelligence to convert legacy code into modern programming languages, and the agency set a goal to finish most of its tech modernization efforts within two years.

Over a year later, the IRS still hasn’t provided the Government Accountability Office with details on its new modernization plan, said David Hinchman, director of IT and cybersecurity at GAO, during a recent roundtable on the IRS.

“Recent changes to IRS' long-term plans have also cast uncertainty over what the agency's modernized end state will look like,” said Hinchman, explaining that the IRS has published “very little” on its new approach.

That’s despite the fact that technological progress is a lynchpin in the IRS’ bigger, overall strategy. After already pushing out over 28,000 employees since Trump took office, the tax agency is aiming to shed more staff and use technology to make up the difference, it said in its recent budget request.

“Without modernization, the IRS would be unable to sustain performance with a reduced headcount,” the budget request said.

Compounding the lack of transparency is a pause in strategic workforce planning, said Hinchman, which would help ensure that the IRS has the right workforce to get the job done.

The IRS didn’t respond to a request for comment.

The IRS hasn’t spared its IT shop from the workforce upheaval that has taken place over the last year. The IRS lost over 2,600 IT employees between January 25 and December 18 of last year, a 31% reduction, according to the National Taxpayer Advocate’s 2025 report.

The IRS also moved over 1,000 IT staff to the office of the chief operating officer last winter, and transferred some of those to jobs helping with filing season, along with human resources specialists.

Agency IT leadership recently told staff that the agency plans to hire 175 IT employees, a tech employee at the agency said.

Even so, IRS leadership isn’t sharing much information internally on the agency’s current plan for its technology, a second tech employee told Nextgov/FCW. Detailed IT strategic plans used to be available within the agency, they said. What’s now available is very abstract.

“It’s hilarious,” they said of the claim that the IRS can use tech to make up for fewer employees.

Using technology to do more with less might be possible in the long run, but “not right now,” the first IT employee told Nextgov/FCW. “We are so short-staffed.”

Despite the staffing shortages, the IRS is planning to capitalize on updates to the online accounts it offers for taxpayers to give Americans access to more self-service options.

The agency is also building a single interface for customer service representatives to allow them to see data about taxpayers that’s currently stored in disparate systems in one centralized place. The IRS thinks this will cut down on call times by speeding up the work of those manning the phone lines. The agency has been trying to build this system since the second Obama administration.

The IRS has also long been working to modernize its core system for individual tax account data, called the individual master file.

The tax agency did put its long-awaited new processing engine for the system into production last year, it said, but more work needs to be done. The effort is one of the most complex modernization efforts in the federal government, and the individual master file touches hundreds of other IRS applications.

“They are shoving AI at us and they think that with that, things can be converted super quickly,” the first employee said of efforts to modernize legacy systems. “But they don’t have the right tools for us yet.”

Frank Bisignano — the head of the Social Security Administration who is also helming a new chief executive officer role at the IRS — told senators earlier this month that data and AI are helping the tax agency with enforcement, even as it’s lost staff and is set to lose more under the Trump administration’s recent budget request.

But GAO reported recently that the agency is facing a skills shortage that could hamper its ability to roll out AI, including systems to prioritize audits.

Among the things that aren’t clear to Congress’ watchdog are how the IRS’ new plans relate to its old strategy to modernize, as well as how and if certain endeavors are continuing, said Hinchman. This isn’t the first time that watchdogs have dinged the IRS for a lack of IT planning.

Leadership under the new administration has altered at least some efforts started during the Biden years. The IRS launched an initiative to digitize paper with funding from the Inflation Reduction Act in 2023 by developing an in-house system. Last spring, leadership directed the IRS to stop working on the project, despite spending $61 million on it already, and shifted to a new approach using contractors, according to a watchdog report.

The IRS also shuttered the Direct File program, launched in 2024 to help certain eligible Americans file their taxes with the government online for free.

]]>

GSA No. 2 talks ‘million hours challenge,’ scaling agency AI efforts

Frank Konkel — Mon, 20 Apr 2026 11:26:00 -0400

The General Services Administration is working to save and automate one million hours of workload across the agency as part of its Eliminate, Optimize and Automate — or EOA — playbook.

GSA Deputy Administrator Michael Lynch said the effort aims to use artificial intelligence and intelligent automation to handle “repetitive, manual workflows,” allowing its workforce to redirect their time toward serving customer agencies, improving procurement outcomes and pursuing other mission-critical services. Early into 2026, Lynch said the agency is almost halfway toward achieving its moonshot goal.

“We have about 400,000 hours that are currently identified of ways that we can — not replace people — but remove that non-high-value added time and replace it by putting people on more high-value opportunities within the agency,” Lynch said April 14 at the OpenText Government Summit in Washington, D.C. “And that really goes to address some of the workforce challenges we’ve had.”

Lynch added that GSA, like most agencies across government, lost headcount over the past year, but he asserted the leaner setup hasn’t slowed operations.

In 2025, the agency took on a centralized role in federal acquisition and led a massive overhaul of the process; became a key cog in the Trump administration’s AI Action Plan through the launch of USAI.gov; and negotiated more than two dozen deals with tech companies, saving partner agencies more than $1 billion on software through its OneGov program, according to Lynch.

Still, the agency is looking at more ways to drive innovation, foster collaboration and develop future leaders within its ranks — while tackling some of the agency’s toughest problems.

Called GSA Labs, the new program seeks a few dozen high-performing early-to-mid-career employees within the agency who will be placed into small, cross-functional teams with executive sponsorship to tackle those problems.

“We did a call to action for all of our senior leaders to say, ‘What are the problems that exist within automation technology, workflows and things that you want to dedicate resources to, that you just don't have the staff to do it?’” Lynch said. “And then we put out a call to our workforce to say, ‘Hey, if you’re kind of mid-career talent, would you be interested in doing a second job in addition to your day job — not a second pay job — but you know, be part of this program?’”

Lynch said the response was “amazing,” totaling more than 300 internal applicants who were narrowed to an initial cohort of 30 GSA staff to address five problem statements with a goal of “trying to create interoperable systems and processes."

He noted GSA Labs’ work is largely internal now, but “could be a scalable program in the future.”

“We have 30 individuals from GSA that are going to be almost our internal McKinsey consulting group that’s going to come in and help us solve these problems in partnership with leaders,” Lynch said. “And then the hope would be that further develops our workforce, develops great outcomes for our agency. And then where the program goes in year two and beyond is a bit up in the air.”

Maturation of OneGov

Far from a one-hit wonder, Lynch said to expect a maturation of GSA’s OneGov effort, which netted more than two dozen discounted deals for agencies from AI and tech software firms, including Google Public Sector, Adobe, Salesforce, Elastic, Oracle, Uber, IBM, Docusign, Microsoft, Amazon Web Services, OpenAI, Anthropic and Box.

Some of those were relatively short-term deals, and GSA is using feedback from industry and agencies to inform how that program evolves into “longer-term, scalable programs and engagements that are mutually beneficial for both the federal government as well as our industry partners.”

“Over the next six to nine months, you’ll see a lot more announcements around how those OneGov deals have matured,” Lynch said.

]]>

Agencies report over 3,000 AI use cases in 2025

Alexandra Kelley — Fri, 17 Apr 2026 08:00:00 -0400

The Office of Management and Budget unveiled the completed 2025 Federal Agency Artificial Intelligence Use Case Inventory, documenting 3,611 individual use cases across 56 submitting agencies.

This represents a 105% increase from 2024’s reported use cases, which came in at 1,757.

Pursuant to several executive actions and congressional laws, including the 2023 National Defense Authorization Act and an April 2025 memo from the Office of Management and Budget, federal agencies are required to report how AI tools are being used in their operations.

OMB is the only office agencies are required to report to, the agency confirmed, adding that use cases employed by the Intelligence Community and Pentagon are exempted from reporting, and other agencies are exempted from reporting National Security use cases. The complete repository was last updated on April 14.

The Department of Health and Human Services reported the most active use cases at 447 — a sizable leap from the 271 reported in 2024. Trailing HHS with the largest number of use cases are NASA, the Department of Veterans Affairs, the Department of Energy, and the Department of Justice, reporting 425, 367, 340 and 314, respectively.

HHS drove the growth in AI acquisition both in 2024 as well as in 2025.

Use case rationales are diverse. Text identification, summarization and manipulation; data analytics and pattern recognition; predictive modeling; and chatbot services were all listed as common use case areas in both high- and low-impact programs.

Of the 445 reported use cases that were deemed “high-impact,” Veterans Affairs reported using the most, with 215 high-impact cases. Some of these include: “Summarization of Clinical Data;” “Forescout: AI Behavioral Anomaly Detection w/ Automated Enforcement;” and “VA VoiceBot for Call Center Modernization.”

AI was also applied in some high-impact use cases to focus on natural language processing for text analysis, machine intelligence for small satellites, science translation and identity verification.

Agencies also listed the software providers they were contracting with to support each of the use cases in their respective inventories. Microsoft’s products dominated the list, with 102 use cases operating with a version of the company’s Copilot AI tool.

Anthropic’s Claude was also a frequently mentioned commercial-off-the-shelf solution used in government AI 25 times, with most use cases coming from the Department of State and HHS. Since the fallout between Anthropic and the Pentagon over the company refusing to allow its AI to be used in surveillance and autonomous weapons, the Trump administration subsequently ordered federal agencies to shed their contracts with the company and stop using its software.

]]>

AI is helping VA speed up claims processing, but Dems worry about errors

Edward Graham — Fri, 17 Apr 2026 07:00:00 -0400

The Department of Veterans Affairs has been using artificial intelligence to speed up its processing of veterans’ benefits claims, but VA officials told lawmakers that human reviewers make the final decisions and that greater use of the emerging capabilities has not correlated with an increase in errors or issues — a claim disputed by Democrats.

During a House Veterans’ Affairs Committee hearing on Wednesday, Margarita Devlin — principal deputy under secretary for benefits at VA’s Veterans Benefits Administration — told lawmakers that “the average time to process a claim has dropped by 42% since January 2025, from 141 days to just 81 days.”

VA’s recently released 2025 inventory of AI use cases listed 367 examples of the emerging capabilities. Twenty-eight of these use cases are focused on the topic of “government benefits processing,” with the majority of these examples listed as still being in the pre-deployment phase.

One of the active AI use cases — and the one that received the most attention during Wednesday’s hearing — is called Automated Decision Support, or ADS.

VA’s inventory said ADS uses machine learning to automate “some of the up-front time-consuming development activities of retrieving information” and noted that the tool “is not intended to replace trained claims processors — it provides tools to assist with development tasks at a time when [the Veterans Benefits Administration] is receiving more claims than ever before.”

Devlin said the tool can review many elements of a veteran’s claim, such as, “if the veteran served in the Navy, were they then within 12 nautical miles of certain areas to qualify for certain presumptive? It'll look for what were their dates of service, and be able to put that all together. So what the employee is seeing is screens that present to them, ‘this has been satisfied, this piece has been satisfied.’”

Rep. Morgan Luttrell, R-Texas, said he was speaking on behalf of the House panel’s general sentiment when he said “we are eagerly behind the injection of AI models into the system so our veterans get what they need in a timely manner.”

Even as lawmakers expressed cautious support for the use of AI and automated capabilities, like ADS, to streamline claims processing, they pressed VA officials to ensure that the tool is working appropriately and includes human oversight.

Devlin said the AI “does not make any decisions and will not deny a claim,” adding that “it simply puts everything together for the decisionmaker so that they can make the decision faster.” She said that the department also regularly updates the ADS system, estimating that it occurs every six to eight weeks.

Sandra Flint — Deputy Under Secretary for Field Operations at VA’s Veterans Benefits Administration — told lawmakers that even if some of a veterans’ collected data appears to be missing from the tool, “that decisionmaker has the opportunity to go out and get the information, or go find the information he or she needs to to support the veterans’ claim.”

Democrats on the panel, however, were more critical of VA’s use of AI and the ADS tool, linking more reliance on these types of capabilities with what they deemed to be decreasing claims-processing accuracy.

Rep. Maxine Dexter, D-Ore., said she had the opportunity to speak with veteran service officers in her district, who told her that decision times were improving but that “their impression is that errors are increasing.”

Dexter also submitted for the record a claim that, she said, explained a veteran’s employability by referencing Google and that was “followed by a quote — direct quote — from language that certainly appears to be AI-generated content.”

Rep. Tim Kennedy, D-N.Y., also questioned the precision of VA’s reported error rate and said “speed does not equal success.”

Devlin told lawmakers that, as of the end of March, the accuracy rate was over 94%, which she said was “the highest it’s been in two years.”

But Kennedy noted that VA reported a “94% issue-level accuracy, but [an] 83.31% claims-based accuracy.”

Flint said the department switched to an issue-based quality rate “because that gives us a level of detail about what's actually happening in the claim, so we can hone in on the things that are important to veterans,” and added that a “claim-based score sort of masks what the real challenges are, because if you get one wrong, the whole thing is wrong.”

In a Wednesday press release, VA noted that it has reduced its backlog of veterans waiting for benefits “to less than 100,000 claims for the first time since 2020.”

These numbers had ballooned as a result of the coronavirus pandemic, as well as following passage of the PACT Act. The law, signed by former President Joe Biden in August 2022, provided veterans exposed to burn pits and other toxic chemicals during their service with expanded access to health services and benefits for medical conditions not previously covered by VA.

In addition to the AI tool used to streamline claims processors’ review of veterans information, VA is looking to expand its use of the emerging capabilities to other benefits processing responsibilities.

VA Press Secretary Peter Kasperowicz told Nextgov/FCW last month that the department “plans to broaden the ADS program to cover more types of conditions and claims,” which he said “means a greater percentage of claims will be eligible for automated processing, speeding up the process and reducing the workload for claims processors.”

VA is also pushing to fund additional adoption and use of AI capabilities.

The Trump administration’s proposed fiscal year 2027 budget, which was released earlier this month, recommended allocating $130 million to the Veterans Benefits Administration “for automation and artificial intelligence investments modernizing veterans claims processing by reducing errors and delivering benefits to veterans faster.”

An FY27 VA budget document focused on the department’s IT operations also proposed boosting its “Decision Intelligence and Automation” activities to $47.8 million, or $4.7 million more than what was enacted in 2026.

The budget document said that “the increase is driven primarily by the AI Infrastructure solution, enabling VA to pilot and scale AI tools that improve operational efficiency, enhance clinical decision-making, and support personalized care and benefits delivery, while sustaining governance frameworks for safe, effective, and ethical use.”

]]>

How public records requests could help ‘fight AI with AI’

Chris Teale — Tue, 14 Apr 2026 18:23:52 -0400

ORLANDO, FLORIDA — Receiving and responding to public records requests is a major task for states and localities, but it has gotten even more complex and time-consuming in recent years.

Governments receive thousands of requests a year, and that number is only set to increase as artificial intelligence supplies automated and repeated requests. Meanwhile, the constraints on staff time, including on attorney’s offices, mean the discovery and redaction processes during and at the end of requests take longer, as they trawl through communications and remove any personally identifiable information and other sensitive records.

But experts suggested AI does not just have to bog down efforts to comply with the Freedom of Information Act and similar state and local laws. Instead, it could help process requests, find relevant documents and take a first go at redaction before humans come in to verify its work.

“We need to fight AI with AI a little bit,” said Erica Olsen, co-founder and CEO of government AI platform Madison AI, during the International City/County Management Association’s Local Government Reimagined Conference in Orlando, Florida, last week. In an interview on the sidelines of the conference, she said public records requests remain the “biggest administrative headache for any agency,” in part due to a reluctance to use tech to help reduce the burden.

An effort to use AI for FOIA requests is already somewhat underway at the federal level. Last year, the National Archives and Records Administration found in a report that almost 20% of agencies who responded to the survey said they use the technology or machine learning in processing requests.

But NARA warned that it will not be as simple as letting technology handle the entire FOIA process.

“AI and machine learning have the potential to aid in FOIA processing but are not a substitute for a FOIA professional’s judgment on application of exemptions and foreseeable harm,” the report says. “It is important that agencies explore the use of AI and/or machine learning options to help improve FOIA processing response times.”

One of the biggest obstacles to timely responses to public records requests is the need to manually review records, something that has become more challenging as the definition of a public record has become broader . Now, governments are not just required to produce emails when asked for records, but also other communications like instant messages.

“The opportunity is absolutely to rethink that whole process and instead of engaging the [government’s] technology services team, fetch [records] through a specifically written AI assistant that fetches all the data, the records, emails, evaluating those records [like] a human is doing,” Olsen said in an interview at the ICMA conference. “But AI can do an initial review of that. It can also do an initial review of what needs to be redacted. And then the human in the loop and the district attorney can do the final review.”

AI can also help refine and validate public records requests, especially if they are too broad or need more information. The technology already helps validate information from the public in areas like professional licensing and has helped speed up that process by checking paperwork and asking for more when needed. A similar initiative could help check and modify FOIA requests.

“Most often, those processes fail at some point because the information that was provided by the resident was incorrect, invalid, or somebody did not qualify,” Luke Norris, vice president for platform strategy and transformation at software company Granicus, said during a session at the ICMA conference. “So now AI can actually help us do that. We're effectively using our staff’s time, and they're not reviewing applications that simply would not be compliant, but instead are actually helping that resident fulfill and create and do that work the first time in the right way.”

Some governments have tried to help ease the burden by suggesting governments raise the costs associated with FOIA requests in a bid to reduce their number. Legislation in California would make the fees associated with requests in the state higher, in an effort to discourage nuisance filers. But experts said there must be another way to make it easier on staff.

“One choice is a charge, that's a choice,” Olsen said. “Or the other choice is to figure out how to use tools to meet the service demand from the community without hiring more staff or charging the community. It's either creating equitable access to information, or not.”

In time, AI could totally transform the public records process and make it something close to self-service with requests fulfilled in a matter of hours or days rather than weeks, months or years, Olsen said. But, she added, there is a long road ahead.

“If we really think about where this could go, that is sort of the Holy Grail of a fully transparent government,” Olsen said. “Some folks are quite scared of that. Clearly, we’ve got a way to go. The data is not ready for that yet. But if we think about the big bold vision, that would be the big bold vision, that any city would have their AI solution such that a citizen can ask for anything they needed, information wise.”

]]>

Treasury is creating a database with pandemic aid recipients’ sensitive information

Natalie Alms — Fri, 10 Apr 2026 07:00:00 -0400

The Treasury Department is pooling information about people who received benefits from pandemic-era relief programs in a new, central database it says will be used to conduct program audits.

It’s the latest front in the Trump administration’s efforts to centralize government data, including information typically held by states about people who receive nutrition benefits and jobless aid. Many of the administration’s previous attempts have been subject to lawsuits.

Critics say the department’s required notice for the system is imprecise, overly broad and runs afoul of privacy laws governing the federal government. Treasury is amassing addresses, financial data, Social Security numbers and other data in the new system, which it says it may cross-match with other government data.

Typically, these types of notices are “routine matters that do not warrant comment,” Steve Sharpe, senior attorney at the National Consumer Law Center, said in a statement. “But the scope of this notice is an astonishing and dramatic departure from prior Treasury practice.”

The NCLC, a nonprofit focused on economic justice, called the new system a “baseless violation of privacy” in a comment on the February notice that it submitted with over 40 other organizations, including many state and local legal aid groups.

Treasury’s plan “could be construed to reach millions of individuals,” the comment reads.

The database will include information about the individuals and entities, like small businesses, receiving benefits from eight department programs, Treasury’s notice says. Congress created many of these during the pandemic to provide emergency relief.

Other programs feeding data into the new system, like one created to rebuild the Gulf Coast after the Deepwater Horizon oil spill in 2010, have no relation to the pandemic. The new system could also include other programs administered by the Treasury in addition to those listed in the formal filing, the notice says.

Local governments administer some of these programs, and they’re already required to report subrecipient and vendor information, the National League of Cities, the United States Conference of Mayors and National Association of Counties say in a comment, which also emphasizes the cost that reporting new data would entail, especially after some of these programs have been shuttered following the end of the pandemic.

Treasury did not respond to a request for comment. But if it moves forward with the new system as described in the notice, it will be saving information about a long list of people — not only those who receive assistance, but also people “associated” with the nonprofits, small businesses and other entities that received or delivered aid. The system will also house application information, which could include sensitive financial information.

The nonprofit Association of Public Data Users wrote in its comment that the notice “seems designed particularly to obfuscate the purpose of the collection and potential uses of the data.”

Although the stated purpose of the system is for audits, “we suspect the unstated purpose of the system of records is not to audit at all, but to get access to the information held by states that Treasury cannot otherwise directly compel them to submit to the federal government,” continues APDU’s comment, which it submitted with nine other organizations.

The Trump administration has pressured a range of state and local entities to share data with the federal government since the beginning of last year, including voter files.

The Electronic Privacy Information Center argues in its comment, which got sign-on from other organizations like the Center for Democracy and Technology, that the new system runs afoul of the Privacy Act’s principles of minimizing data collection, calling the proposed program “illegal and reckless.”

The notice signals more of the same “data grab playbook” from the administration, John Davisson, litigation director for EPIC, told Nextgov/FCW.

“Time and again we've seen this administration exploit personal data to construct wildly exaggerated narratives of waste and fraud, to carry out brutal immigration enforcement tactics, and attempt to undermine the right to vote,” he said.

]]>

CIA employees will get AI 'coworkers'—and eventually run teams of AI agents, deputy says

David DiMolfetta — Thu, 09 Apr 2026 16:59:00 -0400

The Central Intelligence Agency aims to integrate artificial intelligence-powered “coworkers” into analysts’ workflows in the coming years as part of an effort to rapidly adopt the emerging capabilities for use in intelligence-gathering and analysis, a top official said Thursday.

CIA Deputy Director Michael Ellis said these AI coworkers would be housed in agency analytics platforms to help with basic tasks, though humans would still be looped into the process.

“It won’t do the thinking for our analysts, but it will help draft key judgments, edit for clarity and compare drafts against tradecraft standards,” he said in a speech at a Special Competitive Studies Project event focused on AI and the intelligence community. The AI tools would provide triage assistance and flag trends for human analysts to conduct further review.

Within a decade, the CIA will treat AI tools as an “autonomous mission partner” and officers will manage teams of AI agents in a hybrid model to increase the speed and scale of intelligence work, Ellis added.

Last year, the agency had more than 300 AI projects, and, for the first time in its history, AI was recently used to generate an intelligence report, he said.

The remarks provide a rare public glimpse into how one of the nation’s top spy agencies is integrating frontier AI systems into its day-to-day operations, and they signal that such platforms are expected to become a daily feature of officers’ workflows in the near future.

The CIA primarily executes and coordinates human intelligence gathering overseas, often done undercover. Officers recruit and manage foreign assets to clandestinely gather intelligence on areas like economics, terrorism and cyber threats.

Much of that work often involves the use of technology, though some have recently argued the advent of advanced AI tools may push the CIA more toward old-world tradecraft techniques.

But there have been benefits to technological investments. The agency recently elevated its Center for Cyber Intelligence into an entire mission center, a move that’s “paying dividends already by allowing us to deploy new tools to the field and gain more access to priority targets,” Ellis said.

“The battle of cybersecurity will be a battle of artificial intelligence,” and whoever capitalizes on the best AI models will wield “enormous power,” he added. “Having a new mission center centered around cyber intelligence will put us on the path to secure the upper hand.”

The agency also recently announced a new acquisition framework to overhaul how it integrates technology into its missions.

In an effort to track how foreign adversaries like China are using advanced AI and other technologies, the CIA doubled its technology-related foreign intelligence reporting, said Ellis. Those intelligence products focus on technology use abroad and can include findings on areas like semiconductors, cloud computing, infrastructure, cybersecurity or R&D.

Ellis did not mention Anthropic’s recent Project Glasswing announcement, a consortium announced earlier this week meant to help secure critical software against AI-driven attacks. The project was fueled by a powerful, non-public Anthropic frontier model the company says has already uncovered thousands of vulnerabilities but could be weaponized in the wrong hands.

The intelligence community and its industry partners are already examining and discussing how such a model may impact the future of cyber missions, Nextgov/FCW reported Wednesday.

Earlier this year, Anthropic declined to ease restrictions against its tools being used for domestic surveillance or fully autonomous weapons for Pentagon use, triggering a “supply chain risk” designation from the Defense Department and a White House order that all federal agencies phase out their uses of Anthropic tools. The company has legally challenged the move.

Ellis did not single out Anthropic specifically, though he cautioned that the CIA “cannot allow the whims of a single company” to constrain its use of AI and said the agency is looking to diversify across multiple vendors to preserve operational freedom.

]]>