Revolving Door Project | Tracking Uses of AI in the Trump Administration

As the Trump administration continues to expand the use of artificial intelligence (AI) in its operations, the Revolving Door Project has decided to catalogue examples of where and how AI is being deployed. This tracker focuses particularly on uses of AI in the federal executive branch to replace federal workers and undermine transparency and due process.

Government-Wide

New York Times: Trump Taps Palantir To Compile Data On Americans | 5/30/25 

• SUBHEAD: “The Trump administration has expanded Palantir’s work with the government, spreading the company’s technology — which could easily merge data on Americans — throughout agencies.”

• SUMMARY: As Palantir continues to expand its influence within the administration, the Trump administration has given the company the right to surveil  Americans. In a chilling report, The New York Times notes that the company is already creating “detailed portraits of Americans based on government data,” with the Trump administration already seeking “access to hundreds of data points on citizens and others through government databases, including their bank account numbers, the amount of their student debt, their medical claims and any disability status.” 

Immigration

Freedom House: Trump’s Immigration Crackdown Is Built on AI Surveillance and Disregard for Due Process I 5/25/25

• SUBHEAD: “In March, officials at the US State Department revealed that they would use artificial intelligence to revoke the visas of ‘foreign nationals who appear to support Hamas or other designated terror groups.’ The new program, known as ‘Catch and Revoke,’ will scan social media accounts and is part of a broader uptick in the US government’s use of AI-powered surveillance, with the goal of combating antisemitism, terrorism, and illegal immigration.”

• SUMMARY: The State Department’s new policy of stripping away the legal immigration status of foreign nationals for political dissent will ramp up with its new social media surveillance program, “Catch and Revoke.” Under the guise of combatting antisemitism and terrorism, the AI-powered program will sift through the social media data of 33 million people touched by the US immigration system to dissuade First Amendment protected-activity against the Trump administration. The uptick of AI-powered surveillance in the US adds to concerns of rising authoritarianism and the violation of civil liberties. 

404 Media: ICE Taps into Nationwide AI-Enabled Camera Network, Data Shows I 5/27/25

• SUBHEAD: “Data from a license plate-scanning tool that is primarily marketed as a surveillance solution for small towns to combat crimes like car jackings or finding missing people is being used by ICE, according to data reviewed by 404 Media. Local police around the country are performing lookups in Flock’s AI-powered automatic license plate reader (ALPR) system for ‘immigration’ related searches and as part of other ICE investigations, giving federal law enforcement side-door access to a tool that it currently does not have a formal contract for.”

• SUMMARY: Immigration and Customs Enforcement (ICE) is expanding its immigration crackdown by using tools contracted to local law enforcement without publicly disclosing their use. According to data reviewed by 404 Media, more than 4,000 immigration focused license plate searches have already been performed by local and state police on behalf of the federal government, aiding in the Trump administration’s push for mass deportations.

NPR: Former Palantir workers condemn company’s work with Trump administration I 5/25/25

• SUBHEAD: “Thirteen former employees of influential data-mining firm Palantir are condemning the company’s work with the Trump administration weeks after Immigration and Customs Enforcement reached a deal to pay Palantir $30 million to provide the agency with ‘near real-time visibility’ into the movement of migrants in the U.S.”

• SUMMARY: Former Palantir employees wrote a letter in which they spoke out against the company’s partnership with the Trump administration to aid deporting 1 million immigrants this year. The letter’s signees believe Palantir has violated the ethics requirements in its code of conduct, which said its software would protect the vulnerable and develop AI responsibly. The letter also raised concerns about Big Tech’s complicity in undermining democratic principles, calling out Palantir’s violent rhetoric and normalization of authoritarianism. Palantir, cofounded by Trump ally and JD Vance mentor Peter Thiel, has contracts with the US and Israeli militaries, and has seen its stock’s value increase by more than 200 percent since Trump won the 2024 election.

FedScoop: Customs and Border Protection taps ‘chatCBP’ to assist workforce I 5/30/25

• SUBHEAD: “U.S. Customs and Border Protection is implementing an AI chatbot called ‘chatCBP’ for its workforce, following in the footsteps of similar federal government creations like DHSChat and StateChat.”

• Summary: Customs and Border Protection (CBP) will begin using an AI-powered chatbot to provide the same type of generative AI assistance as the popular ChatGPT. The CPB chatbot is touted to improve efficiency and offers features that assist workers complete routine tasks, like document summaries and analysis. Previously, OpenAI’s ChatGPT was used for the same purposes, but the DHS reversed that policy once an in-house tool was developed.

DOGE

ProPublica: DOGE Developed Error-Prone AI Tool to “Munch” Veterans Affairs Contracts  I 6/6/26

• SUBHEAD: “As the Trump administration prepared to cancel contracts at the Department of Veteran Affairs this year, officials turned to a software engineer with no health care or government experience to guide them. The engineer, working for the Department of Government Efficiency, quickly built an artificial intelligence tool to identify which services from private companies were not essential.”

• SUMMARY: DOGE deployed a faulty AI tool developed by Sahil Lavingia to “munch” contracts at the Department of Veteran Affairs. Lavingia’s hastily built tool, which relied on “outdated and inexperienced AI models,” identified over 2,000 contracts for munching, often based on a misread on contract sizes and data.  In fact, the tool wrongly determined that over a thousand contracts, some of which were worth just $35,000, each had a $34 million valuation..Some other “munched” contracts included a gene sequencing device used to develop cancer treatments and a tool used to measure and improve care from nurses.

Australian Financial Review: Musk pushes Grok AI on US government, raising ethics issues I 5/25/25

• SUBHEAD: “Billionaire Elon Musk’s DOGE team is expanding use of his artificial intelligence chatbot Grok in the US federal government to analyse data, said three people familiar with the matter, potentially violating conflict-of-interest laws and putting at risk sensitive information on millions of Americans.”

• SUMMARY: While still in the apparent good graces of President Trump, Musk had DOGE utilize his AI chatbot Grok to analyze government data. According to the Financial Review, the xAI CEO had used a customized version of the company’s chatbot to sift through data at the Department of Homeland Security with no departmental approval, potentially “violat[ing] security and privacy laws.” In 2024, five secretaries of states (Michigan’s Jocelyn Benson, Minnesota’s Steve Simon, Pennsylvania’s Al Schmitt, Washington’s Steve Hobbs, and New Mexico’s Maggie Toulouse Oliver) sent a letter to Musk pleading with the billionaire to make changes to Grok, highlighting its tendency to produce inaccurate information. 

WIRED: DOGE Used a Meta AI Model to Review Emails From Federal Workers I 5/22/25

• SUBHEAD: “Elon Musk’s so-called Department of Government Efficiency (DOGE) used artificial intelligence from Meta’s Llama model to comb through and analyze emails from federal workers.”

• SUMMARY: According to WIRED, DOGE affiliates at OPM were using AI to surveil the emails of federal workers who engaged in conduct that the administration would view as “disloyal.”  “Materials viewed by WIRED show that DOGE affiliates within the Office of Personnel Management (OPM) tested and used Meta’s Llama 2 model to review and classify responses from federal workers to the infamous ‘Fork in the Road’ email that was sent across the government in late January. One of the most egregious examples of DOGE’s (mis)use of AI in this is using it to surveil the emails of government workers to examine conduct that would rebut the Trump agenda.”  While not surprising that fealty is the most important quality to Donald Trump, going after rank-and-file government workers is an escalation of his autocratic politicization of the civil service.  

• According to the Washington Post, DOGE has used AI to analyze programs and spending at the Department of Education. The software is hosted on Microsoft Azure, which in 2024 suffered an unprecedented breach that impacted over 50 executive accounts, to look through spending in the department. 

Pentagon

MIT Technology Review: The Pentagon is gutting the team that tests AI and weapons systems  I 6/10/25

• SUBHEAD: “The Trump administration’s chainsaw approach to federal spending lives on, even as Elon Musk turns on the president. On May 28, Secretary of Defense Pete Hegseth announced he’d be gutting a key office at the Department of Defense responsible for testing and evaluating the safety of weapons and AI systems. As part of a string of moves aimed at ‘reducing bloated bureaucracy and wasteful spending in favor of increased lethality,’ Hegseth cut the size of the Office of the Director of Operational Test and Evaluation in half.”

• SUMMARY: Hegseth’s plan to slash the office that has been described as “the last gate before technology gets to the field” would facilitate department wide use of AI tools and systems that have not been thoroughly evaluated. First established in the 1980s by Congress, the Office of the Director of Operational Test and Evaluation has been pivotal in advocating for the reduction of unknown variables through rigorous testing, providing critical “oversight of the operational testing and evaluation of new systems before a decision is made to begin full-rate production.”

The Register: Pentagon to give AI agents a role in decision making, ops planning I 3/5/25

• SUBHEAD: “The American military has signed a deal with Scale AI to give artificial intelligence, as far as we can tell, its most prominent role in the Western defense sector to date – with AI agents to now be used in planning and operations.”

• SUMMARY: In March 2025, the Pentagon inked a deal with scandal-ridden Scale AI to aid in military planning and operations. Scale AI’s bots will carry out tabletop war-gaming to simulate war plans. Given the widespread issues with inaccuracy and “hallucinations” across generative AI tools, the Pentagon could be relying on low quality and even false data to make complex military readiness decisions. 

SpaceNews: Pentagon boosts budget for Palantir’s AI software in major expansion of Project Maven I 5/22/25

• SUBHEAD: “The Defense Department announced May 21 it is upping the spending limit for software licenses under Maven by $795 million, compared to a ceiling of $480 million just a year ago. The new funding is specifically for U.S. combatant commands, which oversee military operations across geographic regions.”

• SUMMARY: In May 2025, as part of Project Maven, the Pentagon raised its contract ceiling with tech giant Palantir to nearly $1.3 billion through 2029. Palantir has already become deeply entrenched in the Trump administration; in addition to this marked increase in AI use within the military, this also shows the growing influence of the company and its co-founder Peter Thiel.

Energy 

FedScoop:  Energy secretary: Government should ‘get out of the way’ to fuel AI race I 5/8/25

• SUBHEAD: “Energy Secretary Chris Wright views artificial intelligence as ‘the Manhattan Project of our time’ — and the only way to meet the massive energy demands the emerging tech requires is to have the government ‘get out of the way,’ he told lawmakers Wednesday.”

• SUMMARY: Likening AI to the atomic bomb, Energy Secretary Chris Wright said the federal government needs to get out of the way of AI innovation. Speaking before the House Appropriations committee, the former fracking executive claimed that to allow AI to develop, government has to get out of the way so that the industry’s energy needs can be met. The industry’s data centers’ hunger for energy is poised to raise electric bills all over the country, while draining fresh water supply from the places that need it most. 

NextGov: Energy selects 16 sites for AI data center construction, new energy development I 4/3/25

• SUBHEAD: “The Department of Energy is moving forward with some of the Trump administration’s plans to support both artificial intelligence infrastructure development and energy production, identifying 16 public land sites for potential construction.”

• SUMMARY: In April 2025, the Energy Department announced it would move forward with the Trump administration’s plan to offer public lands to build AI data centers. DOE has selected 16 locations that are “are uniquely positioned for the construction of data centers ready to process the large volumes of compute needed for AI applications.” According to Kenza Bryan of the Financial Times, AI data centers are “massively contributing to the continued rise in power demand, which itself contributes to the continued rise in global emissions.” In fact, MIT scientists have estimated the power to maintain data centers went from 2,688 megawatts at the end of 2022 to 5,341 by the end of 2023, projecting the sector’s electricity consumption to reach 1,050 terawatts by the end of 2026. 

Next Gov: OpenAI brings its large language models to Energy’s national labs I 1/31/25

• SUBHEAD: “OpenAI announced on Thursday that it signed an agreement with all 17 of the Department of Energy’s National Laboratories to provide them with access to its reasoning large language models to support breakthrough discoveries for a bevy of applications.”

• SUMMARY: OpenAI inked a deal with the Department of Energy to allow access to its large language models to support the labs’ research efforts. In addition to the immense amount of fresh water that AI data centers use to keep the servers cool, the OpenAI large language model ChatGPT uses about 564 MWh of electricity per day.

FedTech Magazine: What DOE’s FASST Initiative Means for AI Technologies I 5/28/25

SUBHEAD: “The Department of Energy is spearheading an artificial intelligence initiative that aims to ramp up the technology’s use across government. Frontiers in Artificial Intelligence for Science, Security and Technology will rely on DOE’s 17 National Laboratories to build potentially the most powerful integrated scientific AI systems in the world.”

SUMMARY: The Department of Energy is leading a government program that would build one of the most powerful artificial intelligence systems in the world. The Frontiers in Artificial Intelligence for Science, Security and Technology (FAAST) initiative will leverage the data and resources from the DOE’s 17 National Laboratories to enhance the nation’s AI capabilities, with a focus on using the technology to address national security and energy challenges, and drive scientific discovery. 

FAAST has four main goals:

• Transform the DOE’s large datasets into a high quality repository for training AI
• Build AI-enabled supercomputing platforms and infrastructure
• Train and test AI models to predict emergent behaviors while maintaining privacy
• Apply FAAST AI to strategic and critical spaces that lack private sector investment 

New York Times: Energy Dept. Unveils Supercomputer That Merges With A.I. I 5/5/25

• SUBHEAD: “The Department of Energy’s laboratory near the University of California, Berkeley, said it had selected Dell Technologies to deliver its next flagship supercomputer in 2026. The system will use Nvidia chips tailored for A.I. calculations and the simulations common to energy research and other scientific fields.”
• SUMMARY: The Department of Energy announced that its Lawrence Berkeley National Laboratory had chosen Dell Technologies to deliver its new supercomputer. Dubbed the “Doudna,” the new machine will use Nvidia processors for resource intensive tasks like training AI models and genomics research. The DOE’s most recent supercomputer, “El Capitan,” cost upwards of $2 billion to develop and build. The Trump administration has yet to release a funding target or price for the “Doudna,” though its increased specs over its predecessor will likely push the final price tag higher as well. 

GSA

Next Gov: GSA to ‘quadruple’ in size to centralize procurement across the government I 5/3/25

• SUBHEAD: “GSA employees who do still have their jobs will now have access to a new artificial intelligence bot, which was demoed during the Thursday meeting as well. GSA said in an announcement that it plans to offer the tool to other federal agencies, too.”

• SUMMARY: As part of the push to gut the federal civil service, the General Service Administration staff who still have their jobs will now have access to a chatbot intended to replace the work of their fired colleagues. The GSA is an independent agency that is responsible for supporting the basic functions of the federal government, including resource procurement through real estate, technology, and government contracts. 

SSA

Next Gov: SSA is rolling out a new chatbot for employees (and slashing staff) I 5/17/25

• SUBHEAD: “The Social Security Administration is introducing a new chatbot for employees powered by generative artificial intelligence, according to an internal email obtained by Nextgov/FCW.”

• SUMMARY: According to internal plans, SSA intends to roll out a new chatbot to fill the gaps from the thousands of staff cuts. The chatbots are supposed to help with content creation, research, and coding. Notably, the chatbot the agency plans to deploy was not trained with any SSA data, nor was it trained to interact with users. Journalists, working for outlets such as MSNBC and CNET, who have tested the chatbot have come to a similar conclusion: it’s an inefficient tool that leads users to talk in circles. CNET’s Blake Stimac chronicled his experience with the tool, emphasizing its failure to grasp his point about lower payments. 

Newsweek Social Security Announces Major AI Rollout I 3/14/25

• SUBHEAD: “The Social Security Administration (SSA) is implementing a nationwide artificial intelligence (AI) system which it says is aimed at improving the efficiency and accuracy of hearing recordings and transcriptions. The new Hearing Recording and Transcriptions (HeaRT) system is set to replace outdated hardware and is expected to be fully operational by March 17, the agency announced.”

• SUMMARY: In an attempt to apparently modernize its technology, SSA announced its plans to implement  the “Hearing Recording and Transcriptions” (HeaRT)  which seeks to improve the efficiency and accuracy of hearing recording and transcriptions. The system will be used both remotely and in-person. SSA boldly claimed 500,000 customers will benefit from the new system per year. 

ADA: How AI and Technology Are Changing Social Security Disability Claims I 3/21/25

• SUBHEAD: “Additionally, the SSA has experimented with AI-based decision-making to help judges and case reviewers assess claims faster.”

• SUMMARY: Despite some of the flaws with various models, SSA has experimented with using AI to analyze disability claims, identify patterns, and flag applications for further review. Relying on the AI tools to expedite processing still leaves room for continuation of improper denials and more. According to Gianelli & Morris, “if the AI is trained on historical claims data that reflects patterns of improper denials, it may continue to deny valid claims in similar circumstances. Errors in data processing or programming can also lead to incorrect denials, leaving policyholders to bear the consequences of flawed technology.” 

HHS

The Washington Post: White House MAHA report may have garbled science by using AI, experts say I 5/25/25

• SUBHEAD: “Some of the citations that underpin the science in the White House’s sweeping ‘MAHA Report’ appear to have been generated using artificial intelligence, resulting in numerous garbled scientific references and invented studies, AI experts said Thursday.”

• SUMMARY: AI experts said that a report released by the HHS was partially generated by AI, resulting in inaccuracies and even invented studies. The “Make Our Children Healthy Again” report, written to address America’s lagging health outcomes, included 522 footnotes to scientific research, but investigation showed that at least 37 appeared multiple times, 21 hyperlinks were dead, and at least one cited study appeared to be entirely fabricated. Some referenced URLs included the phrase “oaicite,” a definitive sign that the research was collected using the artificial intelligence website OpenAI. The reporting drew criticism from lawmakers and added another black mark to HHS secretary RFK Jr.’s already abysmal record on health science and research.

Healthcare Dive: HHS lays out strategic plan for healthcare AI I 1/14/25

• SUBHEAD: “The HHS has released a strategic plan that lays out the agency’s road map for overseeing artificial intelligence in the healthcare sector.”

• SUMMARY: The plan—which covers AI in medical research, products like drugs and devices, healthcare delivery, social services and public health—comes as AI has become an enticing emerging technology for healthcare executives looking to stretch the industry’s often-beleaguered workforce. The HHS released a strategic plan laying out a road map for AI oversight in healthcare as healthcare executives are turning to AI to supplement the industry’s understaffed and overwhelmed workforce. The plan covers the use of AI in medical research, drug and medical device development, social services, and public health. The main objective is to coordinate a safe public-private framework that improves quality and accessibility of healthcare while mitigating the dangers that inaccurate or biased AI tools pose to patient health. 

Akin Gump: FDA and HHS Appoint AI Chiefs 5/8/25 

• SUBHEAD: “In the last two weeks, the Department of Health and Human Services (HHS) has filled key artificial intelligence (AI) positions at HHS and the U.S. Food and Drug Administration (FDA). These appointments, as well as recent statements from HHS and FDA leadership, align with the Trump administration’s intention to use AI to improve regulatory efficiency at HHS and beyond.”

• SUMMARY: The HHS designated Peter Bowman-Davis as acting chief AI officer to support the use of AI and keep it in line with the Trump administration’s broader AI policies. Bowman-Davis is a Yale undergraduate who has previously worked at venture capital firm Andreessen Horowitz, whose founders endorsed Trump during the 2024 campaign. The Centers for Medicare & Medicaid Services Deputy Administrator noted that the HHS will likely use generative AI to analyze and draft regulations, a policy that has already led to poorly researched reports.

Fierce Healthcare: NIH seeks comment on creation of first AI strategic plan, adds to tranche of AI support in HHS I 6/6/25

• SUBHEAD: “The National Institutes of Health is seeking comment from industry on a new artificial intelligence strategy for the agency.”

• SUMMARY: The National Institutes of Health is seeking input from the healthcare industry on the agency’s AI strategy and on the use of AI in biomedical research, public health, and clinical support. The NIH is also seeking feedback on how best to promote public-private cooperation on AI development and testing, with a particular focus on “data readiness, trust, translation and workforce.” 

FDA

Arstechnica: FDA rushed out agency-wide AI tool—it’s not going well I 6/5/25

• SUBHEAD: “Under the Trump administration, the Food and Drug Administration is eagerly embracing artificial intelligence tools that staff members are reportedly calling rushed, buggy, overhyped, and inaccurate. On Monday, the FDA publicly announced the agency-wide rollout of a large language model (LLM) called Elsa, which is intended to help FDA employees—’from scientific reviewers to investigators.’ The FDA said the generative AI is already being used to ‘accelerate clinical protocol reviews, shorten the time needed for scientific evaluations, and identify high-priority inspection targets.’”

• SUMMARY: The FDA deployed an error prone Deloitte-led large language model (LLM)  called Elsa to aid in scientific review. The FDA’s goals here are to accelerate the research and review process, but despite the millions of dollars Deloitte has paid to develop the software it’s still very much in beta. According to NBC News, when staff tested the tool “with questions about FDA-approved products or other public information, it provided summaries that were either incorrect or only partially accurate.”

Education

White House: Advancing Artificial Intelligence Education For American Youth I 5/23/25

• SUBHEAD:  “To ensure the United States remains a global leader in this technological revolution, we must provide our Nation’s youth with opportunities to cultivate the skills and understanding necessary to use and create the next generation of AI technology.  By fostering AI competency, we will equip our students with the foundational knowledge and skills necessary to adapt to and thrive in an increasingly digital society.”

• SUMMARY: In April 2025, President Trump signed an executive order that directed the Department of Education to provide grants to help educators understand and better instruct students on AI. In June, 68 companies signed the “Pledge to America’s Youth,” a public-private partnership to help fund this initiative. Some of the companies that signed the pledge included; Meta, Adobe, Intel, Google, Amazon, OpenAI, Microsoft, and Salesforce. 

New York Times: Musk’s Staff Proposes Bigger Role for A.I. Bot In Education Department I 2/13/25

• SUBHEAD: “Allies of Elon Musk stationed within the Education Department are considering replacing some contract workers who interact with millions of students and parents annually with an artificial intelligence chat bot, according to internal department documents and communications.”

• SUMMARY: Musk’s DOGE proposed replacing some of the contract workers who work with students and parents with AI chat bots. This is part of the administration’s move to shrink the federal work force. This is particularly noteworthy since these call centers where contractors work field an average of 15,000 questions per day from student borrowers.