OpenAI Flags 'Critical' Cybersecurity Capabilities in Upcoming Astra AI Model

OpenAI has stated that its upcoming Astra artificial intelligence model has demonstrated significant advancements in agentic coding and cybersecurity, leading the company to conclude it cannot currently rule out the model reaching a 'Critical' cybersecurity capability threshold under its Preparedness Framework. This development has prompted OpenAI to take actions including enacting stricter security controls and pausing internal activities, while also collaborating with government agencies and AI safety organizations for testing.