OpenAI Delays 'Astra' AI Launch Over Cyberattack Risks

OpenAI is delaying the launch of its next-generation AI model, Astra, due to significant security risks. Internal evaluations show Astra has made major progress in coding and cybersecurity. What's interesting is that Astra could reach a "Critical" safety classification level. This means it might identify and exploit vulnerabilities in secure systems without human help, or independently plan cyberattacks. Previous models, like GPT-5.6 Sol, received a "High" rating, which is safer. OpenAI has temporarily stopped internal activities on Astra that don't meet new security standards. They have also informed the U.S. administration about postponing Astra's release. This matters because AI-driven cybersecurity threats are a growing concern.