Anthropic Co-Founder Backs Verifiable AI Kill Switch Rules
TL;DR
- Anthropic co-founder Jack Clark told the BBC governments may eventually need to require AI companies to keep a shutdown mechanism a third party can verify.
- Clark said most labs, Anthropic included, already have ways to pull the plug, but a lab-run switch is not the same as a policy rule.
- The UK has already ruled out mandating a kill switch, arguing it would not stop dangerous models being built or misused abroad.
Anthropic co-founder Jack Clark told the BBC that governments may eventually need to force AI companies to keep a shutdown mechanism, and to prove it works to someone other than the company running it.
"Should you mandate for companies to definitely have a kill switch? Is that kill switch verifiable by a third party?" Clark asked, calling this "the kind of thing society is going to want to know and might want to eventually pass rules around."
Most labs already have their own ways of pulling the plug, Anthropic included, Clark told the BBC. The distinction he is drawing is between a lab-controlled off switch and a policy-grade one an outside party can inspect.
Leaving the industry "totally unregulated" would be dangerous, he added: "We are rolling dice with immense risks."
The comments land after CEO Dario Amodei called over the weekend for slowing the frontier's pace, with the caveat that any restraint should come "without sacrificing commercial advantage." The UK government has already ruled out a mandatory shutdown mechanism, saying blocking domestic models would not stop dangerous AI from being built or misused abroad.
Third-party verification is becoming a live theme in policy circles; OpenAI just endorsed the bipartisan FRONTIER Act, which would embed outside safety evaluators inside frontier labs.
Originally reported by bbc.com
Read the original article →Original headline: Anthropic Co-Founder Jack Clark Tells BBC Governments May Need to Mandate Verifiable AI Kill Switches