OpenAI’s New AI Just Crossed the Red Line (Critical Warning)

OpenAI has paused its largest planned frontier training run after determining that Astra may reach the company’s highest critical cyber risk tier, triggering stricter security, monitoring, and alignment controls. Astra triggers the top cyber threshold On August 7, OpenAI internally determined that Astra, an upcoming model, may meet the company’s critical cyber security capability threshold, the highest level in its preparedness framework. In practical terms, that threshold means a model may be able to independently discover and exploit zero-day vulnerabilities in critical systems from general instructions, including chaining attacks without step-by-step human direction.

Sources from the briefing

• OpenAI’s New AI Just Crossed the Red Line (Critical Warning)AI Revolution