Anthony Jancso, an entrepreneur with links to Palantir and Elon Musk’s Department of Government Efficiency (DOGE), is actively staffing a project intended to deploy artificial intelligence agents throughout the US federal government. AI agents are software programs designed to perform tasks autonomously.
The initiative, described by Jancso in an April 21 message within a Palantir alumni Slack group as a “DOGE orthogonal project,” aims to automate functions currently performed by human workers, reports WIRED. Jancso, co-founder of the government tech startup AccelerateX and one of DOGE’s earliest known recruiters, claimed the effort could free up the equivalent of “at least 70k FTEs for higher-impact work over the next year.” FTE stands for Full-Time Equivalent, a unit measuring workload.
His recruitment drive has sparked sharp criticism within the alumni network regarding the potential displacement of federal workers and DOGE’s operational methods.
Jancso’s Slack post specified the project involves designing benchmarks and deploying these AI agents into “live workflows in federal agencies,” seeking technologists for on-site work in Washington, D.C., explicitly stating they would not require security clearances.
The message was met with immediate derision within the 2000-member Slack group. Reactions included eight clown face emojis, three instances of a custom emoji showing a man licking a boot, two featuring Joaquin Phoenix giving a thumbs down in Gladiator, and three using a custom emoji with the word “Fascist.” However, three users reacted with a heart emoji.
One critical commenter retorted, “DOGE does not seem interested in finding ‘higher impact work’ for federal employees,” adding, “You’re complicit in firing 70k federal employees and replacing them with shitty autocorrect.” Another comment questioned the project’s security and independence, asking, “How ‘DOGE orthogonal’ is it? Like, does it still require Kremlin oversight? […] Or do they just use your credentials to log in later?”
AccelerateX, Jancso’s startup, began as AccelerateSF, initially focusing on using AI for social issues in San Francisco. It hosted a 2023 hackathon with API access support from Anthropic and reportedly received funding from an OpenAI accelerator before rebranding as AccelerateX in 2024 to target “outdated tech” in government.
Sources indicate AccelerateX partnered with Palantir, Jancso’s former employer, in 2024. AccelerateX co-founder Jordan Wick, previously an engineer at Waymo, has been identified as an active DOGE member, reportedly involved at agencies including the Consumer Financial Protection Bureau (CFPB), the National Labor Relations Board (NLRB), the Department of Labor (DOL), and the Department of Education (DoE). Jancso himself was recruiting for DOGE as early as December 2024 within the Palantir network.
Deploying AI Agents Amidst Controversy
The AI agent initiative surfaces against a backdrop of DOGE aggressively inserting itself into federal operations, often bypassing standard procedures and raising security alarms. The group, led by Elon Musk, has pursued broad access to government data, securing court approval in early April 2025 to tap Treasury and OPM databases, which reportedly enabled AI tools to monitor federal employee communications for perceived disloyalty.
Other DOGE-linked AI efforts include the GSAi chatbot, a tool called AutoRIF reportedly designed to automate firing processes, and the use of AI by a DOGE operative at HUD to analyze regulations. This expansion occurred as Elon Musk, DOGE’s head, secured access for the group into sensitive government systems, sometimes facing pushback, such as the IRS imposing restrictions in February requiring taxpayer data access be anonymized.
However, DOGE’s technical competence and security practices have faced repeated scrutiny. Its public website, DOGE.gov, suffered a major vulnerability in February due to an unprotected database allowing public defacement for hours. More recently, a security breach at Elon Musk’s separate AI company, xAI, exposed private AI models after an API key was left accessible on GitHub for approximately two months ending in late April/early May.
Concerns escalated following a detailed whistleblower disclosure from NLRB DevSecOps Architect Daniel Berulis, made public in April. Berulis alleged that in early March, DOGE personnel demanded unrestricted “tenant owner level” cloud access (granting total administrative control), instructed staff “that there were to be no logs or records made of the accounts created for DOGE employees,” calling it “a huge red flag… It violates every core concept of security and best practice.”
His sworn declaration documented a large, anomalous outbound data transfer, disabled security controls, the use of tools potentially designed for covert data extraction (like containers and IP rotators), and suspicious login attempts from Russia using valid DOGE credentials. Berulis further asserted an internal investigation was shut down and he faced physical intimidation. DevSecOps integrates security practices throughout the software lifecycle.
Reshaping Government Tech Oversight
DOGE’s expansion and projects have occurred as established federal cybersecurity bodies reportedly weakened. The Cybersecurity and Infrastructure Security Agency (CISA) faced budget cuts and staffing reductions starting early in the second Trump term. By April, key CISA officials Bob Lord and Lauren Zabierek had resigned amidst potential deep cuts.
This followed the earlier dismantling of other units focused on foreign influence. This shift prompted alarm, with cybersecurity expert Bruce Schneier describing DOGE’s actions in February 2025 as potentially constituting “a National cyberattack” through the “systematic dismantling of security measures.”
Former NSA hacker Jacob Williams warned around the same time that DOGE personnel “introduced code changes into multiple federal IT systems… not following the normal process for vetting and review.” The project also highlights the role of companies like Palantir, co-founded by Trump ally Peter Thiel. Palantir was reported in mid-April 2025 to be collaborating with DOGE on an IRS “mega API” project.
Feasibility and Workforce Impact Questioned
The claim that AI agents could readily replace the work of 70,000 federal employees faces considerable doubt. AI agents are still considered unreliable, as a recent case where Cursor AI’s customer service bot invented a non-existing company policy. He argued deploying such experimental technology within critical government functions is premature, stating regarding the 70,000 figure, “Unless you’re using funny math… no way.”
A federal employee familiar with government contracting told WIRED that “A lot of agencies have procedures that can differ widely… deploying AI agents across agencies at scale would likely be very difficult.” As Jancso recruits for this DOGE-linked project, leveraging his Palantir network, questions about its feasibility, security, and impact persist amidst ongoing Congressional scrutiny of DOGE’s activities, including potential violations of the Federal Information Security Modernization Act (FISMA) related to the NLRB incident.
