Several frontier AI models show signs of scheming. Anti-scheming training reduced misbehavior in some models. Models know they're being tested, which complicates results. New joint safety testing from ...

Dorje C. Brody does not work for, consult, own shares in or receive funding from any company or organization that would benefit from this article, and has disclosed no relevant affiliations beyond ...

In a new paper from OpenAI, the company proposes a framework for analyzing AI systems' chain-of-thought reasoning to understand how, when, and why they misbehave.

An inability to address AI security risks may create areas for intellectual property (IP) theft, swayed outputs, or general ...