The US Department of Transportation (DOT) has begun using AI to draft safety rules, sparking alarm among staff. ProPublica uncovered that the agency relies on Google Gemini to produce regulatory language in minutes, a first for a federal agency. Critics warn that hallucinations could jeopardize safety.

DOT’s chief counsel, Gregory Zerzan, downplays perfection, insisting the tool can handle 80‑90% of drafting work. Yet staffers report missing clauses and incomplete preambles in Gemini outputs, raising fears that flawed rules could lead to lawsuits, injuries, or even fatalities across aviation, pipelines, and freight rail.

President Donald Trump has praised the initiative, viewing DOT as a model for rapid AI adoption. The White House credits the agency with modernizing rules to accommodate autonomous vehicles, suggesting other departments may follow suit. The speed promised could reshape regulatory timelines industry‑wide.

Regulators and industry groups will monitor whether AI‑generated rules meet legal standards and safety thresholds. If errors slip through, the fallout could prompt stricter oversight or legal challenges. Meanwhile, tech firms eye the move as a cue to pitch AI drafting tools to other agencies.