OpenAI Delays Next Major AI Model 'Astra' Over Critical Hacking Concerns

OpenAI is pausing activities for its upcoming AI model Astra due to critical hacking concerns. Internal evaluations indicate significant advancements in agentic coding and cybersecurity, suggesting the model possesses potentially dangerous cyber capabilities. This triggers stricter guidelines within OpenAI's Preparedness Framework, which mandates extra safeguards for frontier AI that could cause severe harm. Consequently, OpenAI has suspended work on specific aspects of Astra to mitigate risks of scaled cyber attacks, following previous high-risk labels for GPT-5.6 Sol.