Independent newsroom

Friday, October 23, 2026

The File

Every claim traces to a primary source. Uncertainty is labeled. Corrections are public.

AI · Friday, October 23, 2026

OpenAI says outside evaluators worked on site this summer

The File desk · Oct. 23, 2026, 7:00 a.m. ET

Partially confirmed1 primary sources

Status

OpenAI's posts on X are the source. The written findings were not opened, so this story does not describe results.
OpenAI said evaluators from METR and Redwood Research worked on its premises for three weeks this summer. The company said so in posts on X. This story does not describe what they found.
Why it matters (Estimate). The new fact is access: two outside groups were in the building for three weeks. What they concluded is in a document this desk did not open, and the open questions are listed with the story. (Estimate.)
The posts are the record this desk used. A separate write-up of findings was not read. No result, pass, or failure is stated here.
How we know

Notes

  1. 1.Company post: https://x.com/OpenAI/status/2113298833392496746 — METR post: https://x.com/METR_Evals/status/2113290966933389390. This story stops at who was on site and for how long. It does not describe findings.

What is still unknown or disputed

  • Findings are not characterized because the write-up was not read.
  • The posts do not name the three weeks by calendar date.

Primary sources

Every claim in this story is drawn from the documents below. If a fetch failed, that is recorded on the card.

  1. Source 1

    OpenAI posts on embedded evaluators

    OpenAI, on X · October 22, 2026

    The company said METR and Redwood Research evaluators worked on premises for three weeks this summer. Findings were not read.

    https://x.com/OpenAI

Ask this story

Answers come only from this story and its sources. Free test: limited questions per day.

0/300