The US Department of Transportation is moving toward a faster, AI-assisted rulemaking process, and the plan is drawing sharp concern from inside the agency. According to a ProPublica investigation cited by Ars Technica, DOT has used Google Gemini to help draft safety rules that can affect airplanes, cars, pipelines, and rail systems.
The central tradeoff is speed versus trust. DOT officials see Gemini as a way to compress drafting work that can take weeks or months into a much shorter process. Staffers interviewed by ProPublica worry that even small mistakes in this setting could carry unusually high consequences.
What DOT wants Gemini to do
At a December meeting, DOT’s top lawyer, Gregory Zerzan, described AI as a tool for accelerating rulemaking rather than producing a flawless first draft. The goal, as presented to staffers, was to help the department write rules within 30 days. Zerzan said DOT’s preferred tool, Google Gemini, can draft rules in under 30 minutes.
That framing alarmed some staffers because transportation rules are not ordinary documents. They can shape how safety requirements are written, interpreted, and enforced across systems where errors may have physical consequences.
Zerzan told DOT staffers: “We don’t need the perfect rule on XYZ. We don’t even need a very good rule on XYZ. We want good enough.”
For supporters of the approach, the appeal is clear: AI can produce text quickly and may reduce the time needed to assemble regulatory language. For critics inside the agency, the concern is that speed may hide problems that only experienced specialists can catch.
Why staffers are skeptical
ProPublica granted anonymity to six DOT staffers who discussed concerns about the department’s use of Google Gemini. Staffers said they were “deeply skeptical” that Gemini was suited to this work.
Their objection was not simply that AI can make mistakes. It was that DOT rulemaking is, in their view, “intricate work” that may require decades of “expertise in the subject at hand as well as in existing statutes, regulations, and case law.” That kind of work depends not only on producing fluent text, but on knowing what must be included and why.
ProPublica reported that a Gemini demonstration produced a document that was missing key text, leaving a staffer to fill in the gap. In a low-stakes draft, that might be treated as an ordinary editing problem. In a safety rule, the missing material itself becomes the issue: reviewers must know what is absent before they can correct it.
Some experts who monitor AI use in government told ProPublica that Gemini could save time as a research assistant, if used with supervision and transparency. That is a narrower role than drafting rules at scale. It treats AI output as something to check, not something to depend on as the main engine of regulatory writing.
The safety stakes are broad
The rules at issue are significant because DOT’s work reaches across transportation safety. ProPublica reported that these rules “touch virtually every facet of transportation safety,” including systems that keep “airplanes in the sky,” prevent “gas pipelines from exploding,” and stop “freight trains carrying toxic chemicals from skidding off the rails.”
That breadth explains why staffers are focusing on the possible downstream effects of flawed text. If AI-generated language is incomplete, inaccurate, or poorly matched to existing legal requirements, the consequences may not be limited to an internal drafting problem.
Staffers fear errors could lead to flawed laws, lawsuits, injuries, or even deaths in the transportation system. The concern is heightened by the broader record of AI hallucinations, including cases in which fabricated information has affected courts, lawyers have been fined, and judges have acknowledged that they can be fooled.
One staffer summarized the concern bluntly: “It seems wildly irresponsible.”
Despite those concerns, ProPublica reported that DOT appears to be moving ahead. The department has already used Gemini to draft a “still-unpublished Federal Aviation Administration rule, according to a DOT staffer briefed on the matter.”
How the plan fits into federal AI adoption
Donald Trump has urged federal agencies to adopt AI quickly, according to the source article. ProPublica noted, however, that his orders have not pushed agencies to use AI to draft laws.
Still, Zerzan told staffers that Trump is “very excited” about the DOT initiative. He also suggested that Trump sees DOT as the “point of the spear” and expects other agencies to follow its lead.
The White House has also highlighted DOT’s regulatory changes. In a report on the Office of Science and Technology Policy’s biggest “wins” in 2025, the White House credited DOT with “replacing decades-old rules with flexible, innovation-friendly frameworks,” including fast-tracking rules to allow for more automated vehicles on the roads.
DOT’s expectations for Gemini appear ambitious. ProPublica reported that the department expects Gemini to “handle 80 to 90 percent of the work of writing regulations.” The longer-term vision described in the source is even more automated: federal workers using tools like Gemini would move toward an oversight role, monitoring “AI-to-AI interactions.”
Google’s role and the unanswered questions
Google did not respond to Ars’ request for comment about this use case for Gemini. DOT also did not immediately respond to Ars’ request for comment.
Google has been promoting Gemini for government work more broadly. On Monday, the company posted a blog describing Gemini as a tool that could help federal workers with “creative problem-solving to the most critical aspects of their work.”
The company has also competed aggressively for government AI contracts. According to the source article, Google undercut OpenAI and Anthropic’s $1 deals by offering a year of access to Gemini for $0.47.
Google has described DOT as an important example of its federal work. In a December blog, the company said DOT was “the first cabinet-level agency to fully transition its workforce away from legacy providers to Google Workspace with Gemini.” Google said the move would help DOT “ensure the United States has the safest, most efficient, and modern transportation system in the world.”
Google also encouraged other federal leaders to follow the same path. Its blog said: “We are committed to supporting the DOT’s digital transformation and stand ready to help other federal leaders across the government adopt this blueprint for their own mission successes.”
The unresolved question is not whether AI can draft text. Gemini can produce regulatory language quickly, according to DOT’s own presentation of the tool. The harder question is whether a fast draft is good enough when the document may affect transportation systems where mistakes can create legal exposure, operational confusion, and safety risks.