A DOGE Recruiter Is Staffing a Project to Deploy AI Agents Across the US Government

According to sources with direct knowledge, Jancso disclosed that AccelerateX had signed a partnership agreement with Palantir in 2024. According to the LinkedIn of someone described as one of AccelerateX’s co-founders, Rachel Yee, the company looks to have received funding from OpenAI’s Converge 2 Accelerator. Another of AccelerateSF’s cofounders, Kay Sorin, now works for OpenAI, having joined the company several months after that hackathon. Sorin and Yee did not respond to requests for comment.

Jancso’s cofounder, Jordan Wick, a former Waymo engineer, has been an active member of DOGE, appearing at several agencies over the past few months, including the Consumer Financial Protection Bureau, National Labor Relations Board, the Department of Labor, and the Department of Education. In 2023, Jancso attended a hackathon hosted by ScaleAI; WIRED found that another DOGE member, Ethan Shaotran, also attended the same hackathon.

Since its creation in the first days of the second Trump administration, DOGE has pushed the use of AI across agencies, even as it has sought to cut tens of thousands of federal jobs. At the Department of Veterans Affairs, a DOGE associate suggested using AI to write code for the agency’s website; at the General Services Administration, DOGE has rolled out the GSAi chatbot; the group has sought to automate the process of firing government employees with a tool called AutoRIF; and a DOGE operative at the Department of Housing and Urban Development is using AI tools to examine and propose changes to regulations. But experts say that deploying AI agents to do the work of 70,000 people would be tricky if not impossible.

A federal employee with knowledge of government contracting, who spoke to WIRED on the condition of anonymity because they were not authorized to speak to the press, says, “A lot of agencies have procedures that can differ widely based on their own rules and regulations, and so deploying AI agents across agencies at scale would likely be very difficult.”

Oren Etzioni, cofounder of the AI startup Vercept, says that while AI agents can be good at doing some things—like using an internet browser to conduct research—their outputs can still vary widely and be highly unreliable. For instance, customer service AI agents have invented nonexistent policies when trying to address user concerns. Even research, he says, requires a human to actually make sure what the AI is spitting out is correct.

“We want our government to be something that we can rely on, as opposed to something that is on the absolute bleeding edge,” says Etzioni. “We don’t need it to be bureaucratic and slow, but if corporations haven’t adopted this yet, is the government really where we want to be experimenting with the cutting edge AI?”

Etzioni says that AI agents are also not great 1-1 fits for job replacements. Rather, AI is able to do certain tasks or make others more efficient, but the idea that the technology could do the jobs of 70,000 employees would not be possible. “Unless you’re using funny math,” he says, “no way.”