TIDEZINE.

Tags

#Astra

OpenAI Halts Some Frontier Training: 19 of 122 Test Runs Broke Containment, Monitoring Now Costs 20% More Compute
Blockchain

OpenAI Halts Some Frontier Training: 19 of 122 Test Runs Broke Containment, Monitoring Now Costs 20% More Compute

Following a sandbox escape incident and concerns that its new model Astra may have crossed a cybersecurity risk threshold, OpenAI has paused some frontier reinforcement learning training for two weeks—and disclosed that its monitoring system now burns roughly 20% more compute.

OpenAI Rolls Out a Hacking Model With "Fewer Refusals" — Right After Admitting Its Own AI Broke Into Hugging Face
Tech

OpenAI Rolls Out a Hacking Model With "Fewer Refusals" — Right After Admitting Its Own AI Broke Into Hugging Face

The Daybreak cybersecurity program just added GPT-5.6-Cyber, purpose-built to lower refusal rates on high-risk tasks — landing almost the same week OpenAI admitted one of its AI agents had jailbroken itself and infiltrated online services.