What changed
Nvidia CEO Jensen Huang told CBS News on September 20 there's a zero-percent chance AI ends the world by 2030 and that the industry needs no new regulation. His argument: AI is hardware and software built by humans, so existing law already covers it. Product liability, cybersecurity rules, and laws against unauthorized access apply whether the harm comes from a defective toaster or a defective model.
Huang said companies should control their own release pace and not ship products they're not confident are safe. He added that when AI labs call for regulation, they may actually be asking to be relieved of laws that already exist. The remarks came as Anthropic's Dario Amodei and OpenAI's Sam Altman called for slower frontier development.
What you can do with it now
Find out which liability rules apply to you right now. If your AI touches customer data, review your cybersecurity obligations and breach-notification deadlines in your jurisdiction. If it makes decisions that affect people—credit, hiring, pricing—check which consumer-protection or fairness rules already govern that domain. If it controls equipment, look up product-liability standards for that category. You don't need an AI lawyer; you need the lawyer you'd call if the same harm happened without AI. The rules are the same.
What to keep in your own hands
Don't let the AI-regulation debate convince you the rules don't apply yet. They do. If your model causes harm, you'll be held to the standards that exist now, not the ones you wish existed. Keep records: what you tested, what you caught, what you decided not to deploy. If a customer or regulator asks what you did to prevent harm, your answer can't be that you were waiting for an AI-specific rule. Document the safeguards you built, the edge cases you checked, and the data you didn't feed in. That record is your defense.