AI ‘Kill Switch’ Should Be Mandatory, Says Anthropic Co-Founder
To play this video you need to enable JavaScript in your browser.
AI "more powerful by the day", Anthropic co-founder tells BBC
By Kali Hays and Faisal Islam, Economics editor
Published 23 minutes ago
An artificial intelligence "kill switch" which can be checked by a third party may need to become mandatory for companies, according to Jack Clark, co-founder of one of the world’s leading AI firms, Anthropic.
Clark suggested that society might eventually require rules around such a mechanism to shut off AI software if it becomes too dangerous. He noted that while most labs have ways to pull the plug, policymakers may need to enforce having a universally accepted "kill switch."
These comments come amid growing concerns raised by executives, staff of AI firms, and AI experts who have publicly expressed fears that AI could pose an existential threat to humanity.
Dario Amodei, head of Anthropic, recently called for a slowdown in AI development, but clarified that any actions should not sacrifice commercial advantage. Clark agreed specifics around "kill switch" requirements and verification should be part of the broader policy discussion surrounding AI.
"Should you mandate for companies to definitely have a killswitch? Is that kill switch verifiable by a third party?" Clark questioned. "I think that’s the kind of thing society is going to want to know and might want to eventually pass rules around."
Anthropic, formed in 2021 by former OpenAI employees, is at the center of AI safety debates. Last week, an artificial intelligence researcher who left Anthropic over fears of AI’s potential to wipe out humanity gained widespread attention. In response, Evan Hubinger, a scientist at Anthropic, estimated the risk of human extinction from AI as
">10% within the next decade"
.
Geoffrey Hinton, a computer scientist and Nobel Prize winner known as the "Godfather of AI," agreed with this assessment, telling the BBC that a 10% chance was "not unreasonable." Clark, however, downplayed the usefulness of such statistics but emphasized the need to change the course of the AI industry.
Related Stories:
- UK government rejects ‘kill switch’ idea for dangerous AI (Published 3 days ago)
- US lawmakers propose ‘Kill Switch Act’ for problematic AI tools
- Donald Trump dismisses AI safety concerns as a "HOAX"
