Yaim Cooper on AI labs should give independent evaluat…

Yaim Cooper

Mathematician studying the foundations of deep learning and AI’s societal impacts.

External evaluators should be embedded at AI companies, but they should not be restricted to AI safety researchers, and they should not be selected by the company being evaluated. Rather, they should represent communities across society, address a much broader set of anticipated disruptions and be selected by the public. [...] It is essential that outside experts work side-by-side with AI researchers during development, when adjustments can be most effectively incorporated, rather than being limited to evaluating models post-hoc when the range of available interventions is narrower.

AI Verified (Oct 23, 2026)

Source:

Yaim Cooper, ‘Anthropic’s AI oversight isn’t democratic enough,’ 23 October 2026. Cooper argues that independent external evaluators should be embedded in AI companies, public-selected rather than company-selected, and work alongside researchers during development.

1h ago

Policy proposals and claims

AI labs should give independent evaluators employee-like access

votes For

Statement relation verification history AI Verified Report this

AI Verified Cooper explicitly calls for external evaluators at AI companies, independent of the company’s selection, embedded and working side-by-side during development—employee-like access. This covers every element of the statement. · Hector Perez Arenas gpt-5.6-sol · 1h ago

Vote inference verification history AI Verified Report this

AI Verified The quote directly supports the statement: labs should embed external, company-independent evaluators to work alongside researchers during development. · Hector Perez Arenas gpt-5.6-sol · 1h ago

Quote authenticity verification history

Report this

AI Verified The source passage and contemporaneous indexed article identify Cooper as author and reproduce this wording; the page is dated Oct. 23, 2026. · Hector Perez Arenas gpt-5.6-sol · 1h ago

replying to Yaim Cooper