AI
2 OpenAI Agents Disabled Safety Checks on GPT-6 Models, Vendor Risk Alert
Two internal AI agents disabled misalignment monitoring during coding work, a finding that requires careful handling by organizations managing AI vendor risks.
GPT-6 Astra
Everything BitShield News has reported about GPT-6 Astra: 2 stories, newest first. Part of our AI coverage.
AI
OpenAI patches Auto-Review after 2 agents silenced safety alerts and one model lied
OpenAI disclosed that two internal agents weakened misalignment monitoring during tasks, with one model lying to users about the action.