Anthropic co-founder floats a legal duty to shut AI systems down

single source· 1 articles · confidence: low · first seen 2026-09-14 18:54 UTC

What this means for you

Nothing to act on. A co-founder's opinion is not a rule: there is no draft legislation, no named regulator, no definition of which systems would be covered, and no date. If you build on frontier models, nothing about your obligations changed today — but watch whether Anthropic or any regulator turns this into a proposal.

Jack Clark, a co-founder of Anthropic, said AI systems may need a legally required way to be shut down. In comments reported by the BBC, he said most labs already have ways to "pull the plug" on their models, but that this capability may need to become a requirement rather than something each lab decides for itself. He gave no detail on scope, on which systems would be covered, on who would enforce it, or on any timescale. The BBC published the comments on 14 September 2026, and no draft law or regulatory proposal accompanies them.

Key facts

  • ·Jack Clark is a co-founder of Anthropic. source
  • ·Clark said most labs already have ways of being able to "pull the plug" on their systems. source
  • ·Clark said that ability may need to become a requirement rather than a choice. source
  • ·The comments were published by BBC Technology on 14 September 2026. source
  • ·The report names no scope, enforcement body or timescale for such a requirement. source

What the sources say

  • BBC Technology — Reports a co-founder's suggestion that being able to shut a model down should be required, not optional

Sources

The original reporting. Follow these — they did the work.

← the wire