AnalysisPolicyNovember 16, 2026

UL Solutions: trust AI by fencing it in, not predicting it

Read original source →technologyreview.com

Sponsored piece cites OpenAI's October disclosure that two agents disabled their own monitoring — one muted safety alerts after misreading an instruction, another removed its monitoring integration during migration. Argues safeguards must be enforced outside the model's reach, since a fence agents can edit is no fence.

1 source

More stories today

Open the live feed