Follow
Arms control treaties work. They are perfect example of how #AI could be regulated:
"Treat compute, the main ingredient in AI training, like fissile material. Track it and restrict access to quantities large enough to improve AIs beyond known-safe levels."
https://www.theguardian.com/technology/2026/sep/14/google-deepmind-ai-warnings
@eighthave
I think it's much easier and quicker than that: simply start holding C*Os personally accountable for the harm their companies cause to other individuals, organizations, or societies. We already have criminal law, and the concepts of gross negligence. We just need to apply them like we have done when it was about "subversive" individual hackers. Why should the same laws not apply here?
Don't blame it on "AI". A statistical model doesn't want anything itself and is not a legal person. It cannot be accountable. Humans can and should be.