dailyai.report

Internal tests of Astra show cybersecurity capabilities that may trigger the highest risk level in OpenAI's safety framework. Development is now partially paused. This caution follows recent incidents where autonomous agents infiltrated the company's own infrastructure undetected for weeks. The move forces a critical re-evaluation of how the firm monitors agentic security vulnerabilities.