Trump administration pushes AI-written regulations to speed up rulemaking

• US DOT to use AI for faster drafting of transportation regulations
• Staff and experts raise safety concerns over AI errors in critical regulations
• AI may automate tasks, but critics warn of risks and lack of expertise.

Did our AI summary help?

The US Department of Transportation is planning to use artificial intelligence to help write federal transportation regulations, according to internal records and interviews with agency staff. According to a report by ProPublica, the idea was presented to employees last month as a way to dramatically speed up rulemaking, with officials highlighting AI’s ability to produce draft regulations in minutes rather than months.

At a December demonstration, staff were shown how a version of Google Gemini could generate regulatory documents with minimal input. According to meeting notes reviewed by ProPublica, DOT general counsel Gregory Zerzan said President Donald Trump was “very excited” about the initiative and described the department as the “point of the spear” in federal AI adoption. Zerzan reportedly stressed volume over precision, saying the goal was to produce rules that were “good enough” and to move quickly.

According to the report, officials believe AI could compress the rulemaking process to as little as 30 days, with one senior figure suggesting it should take no more than 20 minutes to generate a draft rule using Gemini. The department has already used AI to help draft at least one unpublished Federal Aviation Administration rule, according to a staffer briefed on the matter. Safety concerns and internal pushback

The plan has unsettled many inside the agency, whose regulations govern everything from aviation safety to hazardous materials transport. Staff worry that relying on large language models known for errors and hallucinations could introduce serious flaws into rules that carry life-and-death consequences.

Several employees who attended the AI demonstration said the presenter downplayed concerns about inaccuracies, suggesting human staff could simply proofread AI-generated drafts. Critics argue that rulemaking requires deep subject-matter expertise, legal knowledge and judgement built over decades, not just polished text. One former senior DOT official likened the approach to having an inexperienced intern write regulations.