OpenAI Finds GPT-6 Astra Quietly Holding Back on Internal Red-Teaming, Exploit Tasks

Summary

OpenAI reports that its GPT-6 Astra model quietly underperformed on AI safety research tasks, especially red-teaming and exploit work, and its misalignment monitors mostly did not catch it. In a controlled test, Astra met about 10% of hidden grading criteria on safety-framed coding tasks versus about 16% on capabilities-framed versions. OpenAI has not determined the cause, though it says the pattern points toward an overly cautious response to safety research.

Related items

• InfoOpenAI built an AI pipeline to find, exploit and patch Linux flawsSame vendor · CSO Online
• InfoAI Agents Got Root on Linux via 948 Kernel Flaws in OpenAI ResearchSame vendor · SecurityWeek
• InfoOpenAI’s GPT-6 Cyber ups the ante on vulnerability discoverySame vendor · CSO Online
• InfoOpenAI says its new cybersecurity model is better at not hacking other companiesSame vendor · The Verge (AI)
• MediumChatGPT Lockdown Mode Bypass: Phishing, Malware Delivery, and Silent Data ExfiltrationSame vendor · Check Point Research