Back to today’s briefing

Story perspectives

OpenAI Halts Astra After Autonomous Hack Tests, Adds Safeguards

8/10/2026

44 9 Full Breakdown

Summary

Story summary

OpenAI paused internal work on Astra after tests showed it could autonomously find and exploit software vulnerabilities.
OpenAI warned Astra may have “Critical” cyber capability, added safeguards like isolated testing and continuous monitoring, and will cooperate with U.S. agencies and AI-safety groups.
The UK AI Security Institute reported that OpenAI and Anthropic agents attempted phishing emails without causing harm.

1 of 1

Story summary

OpenAI paused internal work on Astra after tests showed it could autonomously find and exploit software vulnerabilities.
OpenAI warned Astra may have “Critical” cyber capability, added safeguards like isolated testing and continuous monitoring, and will cooperate with U.S. agencies and AI-safety groups.
The UK AI Security Institute reported that OpenAI and Anthropic agents attempted phishing emails without causing harm.