Beth Barnes, CEO of METR, a group that evaluates the risks of frontier models, says that while embedded at OpenAI, METR could talk only to employees OpenAI approved, and could not quote or summarize what they said.

This is the account of a party to the agreement, describing narrower access than an OpenAI employee had described publicly, and whether the other three outside evaluators OpenAI has embedded got the same terms isn't known. Barnes was correcting OpenAI employee roon on X; roon had thought evaluators could talk privately to any employee. Barnes wrote that METR could speak only to employees the company approved, and could not quote or summarize even the information those approved employees gave (Beth Barnes, 2026-11-17). SecondSource's November 17 item "OpenAI lets four outside evaluators examine models still in training" already noted that what employees say in interviews may not be quoted publicly. METR now states on the record that OpenAI also approves who can be interviewed. How we checked →
Source and checks
This item ran in the brief of Nov 18, 2026, alongside 14 other items. How we checked it, the quote checks and what was still missing that day are on that issue's How we checked page.
Cite this item
SecondSource (2026). "Beth Barnes, CEO of METR, a group that evaluates the risks of frontier models, says that while embedded at OpenAI, METR could talk only to employees OpenAI approved, and could not quote or summarize what they said.." https://secondsource.io/en/expert/2026-11-18-1/. Accessed access date.