ACN · Presented by Levr
ACN WIRE

ACN Wirevia TechCrunch

OpenAI to embed outside researchers to look for AI subversion

Written with AI assistance, reviewed by the ACN desk.

OpenAI will embed safety evaluators from Apollo Research, Goodfire, METR and Transluce in a pilot that examines whether its models try to undermine their own alignment or control mechanisms, TechCrunch reported. OpenAI called the pilot an ambitious first step towards deeper access and disclosure rights. Apollo said true independence requires all data and records relevant to its assessment, which it said it does not yet have.

Full reportTechCrunchPublished Nov 16, 2026Read the full report (opens in a new tab)

Spot an error? Email corrections@aicodingnews.si. Fixes are logged on the corrections page.

Sponsor

All stories →

More from the wire

All wire briefs →

From our sponsors

Sponsor

Sponsor

Sponsor

Sponsor