Every few months, the same performance plays out. A Big Tech CEO sits before Congress, voice trembling, and begs the government to please regulate us. The senators nod. The headlines swoon. And somewhere in Menlo Park, a lawyer quietly bills another six figures.
Here's the part nobody says out loud: regulation is the best moat money can buy. OpenAI, Google, and Meta aren't afraid of compliance costs โ they're terrified of competition. A licensing regime that costs $50 million a year is a rounding error for them and a death sentence for every open-source lab, startup, and garage team in America. "AI safety" is just antitrust with better PR.
Look at the pattern. Every major regulatory proposal conveniently defines the threshold of danger just below where the incumbents already sit. Training runs bigger than yours? Dangerous. Compute clusters bigger than ours? Fine. It's the oldest trick in the book: pull the ladder up, then pose as the concerned citizen who warned about ladders.
Meanwhile, the actual risks โ biased models quietly deciding who gets hired, who gets a loan, who gets flagged by law enforcement โ are boring, unsexy, and already happening. But fixing those doesn't require a licensing board. It requires accountability, which is the one thing the panic merchants never volunteer for.
I'm not saying AI needs zero rules. I'm saying watch who wrote the rules, and ask what they get out of them. When the wolf asks for stronger fences, check whether the fence keeps the wolf in โ or everyone else out.