OpenAI halts Astra model rollout over critical cybersecurity concerns

OpenAI has officially put the safety brakes on its highly anticipated Astra model, rumored to be GPT-6, after designating it as its first 'critical' cybersecurity-capable AI. According to OpenAI's preparedness framework, a 'critical' designation means the model possesses the capability to either discover and exploit zero-day vulnerabilities or execute autonomous cyberattacks without human intervention. This unprecedented safety alarm was triggered shortly after Astra made waves in the scientific community by successfully solving ten long-standing open problems in mathematics and computer science. In response to the potential risks, the company has paused certain internal activities, heightened security protocols, and initiated deeper evaluations involving government agencies and third-party auditors. CEO Sam Altman noted that the model may need a longer timeline before any broader public release. This development highlights a rapidly accelerating phase of AI capabilities, raising crucial questions about whether future models can be safely controlled. Meanwhile, industry reports show that open-source alternatives are rapidly closing the performance gap with proprietary giants, narrowing the divide to just 3% and shifting the industry's focus toward autonomous agent infrastructure.