AI Drafts US Transportation Rules: Safety at Risk?

The US Department of Transportation (DOT) has begun using artificial intelligence, specifically Google Gemini, to draft safety regulations for crucial transportation systems including airplanes, cars, and pipelines. This initiative, revealed recently, has sparked worry among DOT staffers and external experts due to AI's known tendency to generate inaccurate or fabricated information. Concerns include the potential for flawed laws resulting from unchecked AI errors, which could endanger public safety and lead to legal challenges.

Despite these apprehension, DOT's top lawyer, Gregory Zerzan, has indicated that the primary goal is to expedite the rule-making process. He stated that the aim is not for perfect regulations but for 'good enough' ones, with Gemini capable of drafting rules in under 30 minutes, significantly faster than the traditional weeks or months. This approach has been met with skepticism by staffers who emphasize the intricate nature of regulatory work requiring deep expertise.

DOT staff have expressed deep skepticism, noting that a demonstration of Gemini's capability produced an incomplete document requiring significant manual correction. The department's accelerated adoption of AI for rule-drafting follows a period where AI hallucinations have caused issues in legal proceedings. The rules in question are critical for ensuring transportation safety, covering aspects from aviation to the transport of hazardous materials.

Former President Donald Trump reportedly expressed excitement about the DOT's initiative, viewing it as a potential model for other federal agencies. The DOT anticipates that Gemini could handle a substantial portion of regulation writing, potentially reducing human roles to oversight. Google, meanwhile, has been actively promoting Gemini for government use, competing for contracts by offering competitive pricing, and celebrating DOT's adoption as a milestone in digital transformation.