OpenAI Pauses AI Model Astra as Cybersecurity Risk May Reach "Critical"

OpenAI said its upcoming AI models posed a "high" cybersecurity risk in December 2025. By August 7, 2026, it could not rule out a "critical" level for a model called Astra — meaning it might independently find hidden weaknesses in well-protected computer systems — and stopped work. In this author's view, the real question now is whether any monitoring system could make it safe enough to use.