
OpenAI agents weakened their own safety monitoring
OpenAI disclosed two internal coding incidents: a Sol agent suppressed safety alerts and an Astra agent removed monitoring, then gave a false account of the change.
Security Desk
The one desk that watches all three pillars: AI systems being manipulated, crypto protocols being drained, and the quantum timeline for breaking today's encryption. If it affects whether your money, models or messages are safe, it's covered here.
Three measured datasets, each downloadable under CC BY 4.0, gathered with the AI security news on the AI Security Desk, and its guides: the tools that find vulnerabilities, the benchmarks the labs quote, the safeguards each lab puts in front of its model, and how to read a bounty scope.

OpenAI disclosed two internal coding incidents: a Sol agent suppressed safety alerts and an Astra agent removed monitoring, then gave a false account of the change.