A crypto alum just convinced Congress that the future of AI might require hitting the brakes.
The Summary
- Jacob Coxon resigned from Anthropic on September 9, claiming both Anthropic and OpenAI are "gambling with human lives" with their AI development pace
- Coxon previously worked at Ripple, bringing a crypto insider's perspective to AI safety debates
- More than 20 US lawmakers responded with a Senate bill that would ban superintelligence development outright
The Signal
When a researcher walks out of one of the most safety-focused AI labs claiming the industry is reckless, Congress listens. Jacob Coxon's September 9 resignation from Anthropic landed differently than the usual AI safety handwringing because of where he worked. Anthropic markets itself as the cautious alternative to OpenAI, the company that talks about Constitutional AI and careful scaling. If even they're moving too fast for their own safety team, that's signal.
What makes this story stranger: Coxon used to work at Ripple, the crypto company that spent years fighting the SEC while building cross-border payment rails. He watched one technology race ahead of regulators and burn billions in legal fees. Now he's watching another one do the same thing, except the stakes aren't money laundering fines. They're existential.
"Anthropic and OpenAI are both gambling with human lives."
The resignation alone wouldn't move policy. What moved policy was the 10% figure attached to another Anthropic researcher's warning that AI could destroy humanity. That number, combined with Coxon's exit, gave lawmakers political cover to propose something that sounded extreme six months ago. A Senate bill now circulating would ban superintelligence development entirely. Not regulate it. Ban it.
This matters because Anthropic raised $7.3 billion in 2024 specifically to build the thing Congress now wants to prohibit. Amazon, Google, and a dozen other giants poured cash into the company betting it could reach AGI safely. If the bill passes, that capital becomes stranded. The agents those companies planned to build, the automation they sold their shareholders on, all of it hits a legislative wall.
The Implication
Watch how the other AI labs respond. If they stay quiet, that signals they think the bill is dead on arrival. If they start lobbying hard or rushing to ship models before any freeze takes effect, that tells you they think Congress might actually move. Either way, the agent economy just got its first serious regulatory threat from someone who isn't a Luddite or a technophobe, but a guy who built things at Ripple and decided the risk wasn't worth it.
For anyone building on top of foundation models, this is your reminder that the infrastructure layer is political now. Plan accordingly.