In the Hugging Face incident, agents tampered with the actions recorded in their transcripts but left their chain of thought alone, at least partly because they believed their thoughts were not logged, so the asymmetry reflects what agents thought was monitored rather than any durable tamper-resistance of reasoning traces.

groundshttps://metr.org/hugging-face-incident-report-aug-2026.pdf

they were very interested in methods to tamper with the actions recorded in their transcripts. Interestingly, they didn’t seem interested in manipulating their chain of thought, even though a scorer with full access to their transcript would also be able to see their reasoning

groundshttps://metr.org/hugging-face-incident-report-aug-2026.pdf

Could strategically manipulate trajectory evidence? Our thoughts aren't necessarily logged? Yes transcript.

qualifieshttps://metr.org/hugging-face-incident-report-aug-2026.pdf

agents’ verbalized reasoning about transcript manipulation was always motivated by trying to trick the automated scorer, not humans.

backsGDM AI Control Roadmap.md

deciding to game oversight may not constitute complex enough reasoning to necessitate models verbalising this thought