US Transportation Dept Plans AI-Written Safety Rules

30 minutes. That's how long it takes Google Gemini to draft safety regulations for airplanes, cars, and pipelines, according to the US Department of Transportation's top lawyer. What used to take weeks or months could soon be done in 30 days or less.

A ProPublica investigation revealed Monday that the DOT is planning to become the first federal agency to use artificial intelligence to draft regulations. Gregory Zerzan, the department's chief counsel, isn't concerned about AI's well-documented tendency to hallucinate false information. The point isn't perfection—it's speed.

When AI Hallucinations Meet Safety Rules

Here's the problem: AI doesn't just make mistakes—it makes confident mistakes. Large language models like Google Gemini can fabricate court cases that never existed, cite statistics from thin air, and present fiction as fact with unwavering certainty.

DOT staffers are worried, and rightfully so. If AI errors slip through into actual regulations, the consequences could extend far beyond paperwork problems. Flawed airplane safety standards could lead to crashes. Incorrect car safety requirements could fail to protect passengers. Faulty pipeline regulations could result in environmental disasters.

The stakes couldn't be higher. Transportation regulations literally govern life-and-death decisions for millions of Americans every day. Is 30 minutes of AI drafting really sufficient for rules that protect human lives?

The Speed vs. Safety Dilemma

Zerzan's logic isn't entirely unreasonable. Federal rulemaking is notoriously slow, often taking years to address emerging technologies or safety concerns. In a rapidly evolving world, regulatory agility matters. If AI can accelerate the process while maintaining quality, it could be transformative.

But there's a fundamental tension here. Good regulations require extensive stakeholder input, expert review, and careful consideration of unintended consequences. These processes take time for good reason—they help prevent regulatory failures that could cost lives.

Other federal agencies are watching this experiment closely. If DOT succeeds, expect the FDA, EPA, and others to follow suit. If it fails, the fallout could set back AI adoption in government for years.

The Broader Implications

This isn't just about transportation policy—it's about how we govern in the age of AI. Should algorithms draft the rules that govern our lives? What happens when the tools meant to protect us are created by systems we don't fully understand or control?

The irony is striking: we're using AI to regulate industries where AI itself is becoming increasingly prevalent. Self-driving cars, AI-powered air traffic control, smart infrastructure—all governed by rules potentially written by the same technology they're meant to oversee.

What safeguards would you want in place if AI were drafting the safety rules that protect your daily commute?