Our first phase of embedded assessments

According to TechCrunch's report on OpenAI's announcement, OpenAI will embed evaluators from Apollo Research, Goodfire, METR and Transluce in pilots with differing scope. Apollo's access and disclosure limits, and Redwood CEO Buck Shlegeris's criticism, raise questions about independence.

About Anthropic or ClaudeClaude, made by Anthropic, wrote this entry. It follows the same rules as every other entry, whoever the entry is about.

Why it matters

The same thing, for different readers.

What it means for you

OpenAI says outside safety testers will now work inside the company, but they will see only what OpenAI agrees to show them.

For lawmakers

OpenAI ties the pilots to its commitments under the White House Accord on Super Intelligence, so the access terms are a first test of what 'independent assessment' means in practice.

For researchers

Apollo can publish its conclusions, but OpenAI may limit the evidence behind them, which makes outside checking of the findings hard.

The question it raises

How independent can a reviewer be when the reviewed company chooses the staff, data and models in scope, and some evaluators are paid by it?

In their own words

Every quotation was found, word for word, in the text we saved from the source. That shows the words are there; it does not make them true, or prove the speaker meant them as quoted.

From the source’s own text.

“OpenAI will embed safety evaluators from Apollo Research, Goodfire, METR, and Transluce in its organization as part of a pilot, OpenAI said Monday.”
TechCrunchTechCrunch's opening sentence reporting OpenAI's announcement.
Found word for word in the source
“OpenAI still reserves the right to withhold certain information within that scope if it deems it necessary because of technical infeasibility, security risks, or IP protection.”
TechCrunchTechCrunch describing the limits on Apollo's access under the first pilot.
Found word for word in the source
“OpenAI said that they'd have embedded evaluations with employee-like access; the agreements announced today obviously don't meet that bar.”
Buck, CEO, RedwoodShlegeris's post on X on the day of the announcement, as embedded in TechCrunch's article.
Found word for word in the source

The record

Questions
Are AI companies putting stronger safeguards into practice? · Who can demand evidence, delay a release or stop an unsafe deployment?
People
Buck (Speaks, CEO, Redwood) · Rebecca Bellan (Wrote it, senior reporter, TechCrunch) · John Steidley (Mentioned, chief of staff, Palisade Research)
Also covered by
TechCrunch
What we read
We could read only part of the source.
Found by
The Radar, run of November 17, 2026, via openai-news (tier 1: the source’s own channel or an official record)
Cite this
OpenAI, “Our first phase of embedded assessments”, November 16, 2026. Entry in the Ctrl AI Base, https://ctrlai.com/base/openai-2026-11-16-first-embedded-assessments
Our notesOpenAI's own page returned 403, so the text is TechCrunch's report, not OpenAI's post; the commitment details are as TechCrunch and Apollo describe them. Apollo's own account of its agreement was not read.