Stuart J. Russell on AI labs should give independent evaluat…

Stuart J. Russell

AI Expert and Professor

Frontier AI companies should grant embedded evaluators access equivalent to that of their own highly privileged employees for the purposes of their evaluations, and with exceptions to protect sensitive data belonging to the company’s customers and other third parties. This includes access to the same relevant systems, data, tools, and physical spaces as those available to senior internal company employees responsible for carrying out comparable risk assessments, as well as candid and direct one-on-one communication with relevant staff.

AI Verified (Sep 18, 2026)

Source:

Stuart J. Russell, personal-capacity signatory, “Minimum Conditions for Embedding Evaluators,” AI Evaluator Forum public letter, September 18, 2026, conditions 5 and explanatory paragraph. The page identifies Russell among signatories and states that all sign in their personal capacity.

20d 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 The quote expressly says frontier AI companies should give embedded evaluators access equivalent to highly privileged employees, including relevant systems, data, tools, spaces, and staff communication; this covers employee-like evaluator access. · Hector Perez Arenas gpt-5.6-codex · 20d ago

Vote inference verification history AI Verified Report this

AI Verified The quote explicitly supports employee-equivalent evaluator access; recorded answer ‘for’ matches. · Hector Perez Arenas gpt-5.6-codex · 20d ago

Quote authenticity verification history

Report this

AI Verified AI Evaluator Forum letter (Sep. 18, 2026), §§5 and explanatory paragraph, reproduces this wording and lists Stuart Russell as a personal-capacity signatory. · Hector Perez Arenas gpt-5.6-codex · 20d ago

replying to Stuart J. Russell

Other opinions from Stuart J. Russell

See all

Stuart J. Russell votes For

Pausing AI at human-genius level to solve alignment is safer than either racing to superintelligence or halting entirely

This is not a ban or even a moratorium in the usual sense. It’s simply a proposal to require adequate safety measures for a technology that, according to its developers, has a significant chance to cause human extinction. Is that too much to ask?

AI Verified source (Mar 27, 2026)

added 2mo ago

Stuart J. Russell votes For

Ban superintelligence development until safety consensus is reached

This is not a ban or even a moratorium in the usual sense. It’s simply a proposal to require adequate safety measures for a technology that, according to its developers, has a significant chance to cause human extinction. Is that too much to ask?

AI Verified source (Mar 27, 2026)

added 2mo ago

Stuart J. Russell votes For

Mandate big AI companies spend 30% of compute on AI safety

"I've spoken to engineers in those companies who think it's now 60 or 70%," Russell said, referring to the probability of catastrophic outcomes from AI development. "So we're off by a factor of 10 to 50 million. And so that's not a very impressive en...

more Unverifiable source (Feb 19, 2026)

added 5mo ago