OpenAI Astra Paused: Cybersecurity Threat Discovered

OpenAI has paused parts of its new Astra model development. This is because an internal review found the model reached a "critical cybersecurity threshold." This means it could independently perform cyberattacks on real-world systems. This finding triggered new safeguards under OpenAI's Preparedness Framework, established in 2023. The company says its evaluations are ongoing, and a "Critical" capability level can't be ruled out. OpenAI clarified that Astra was not involved in a separate breach of Hugging Face's systems. What's interesting is this disclosure comes amid increasing scrutiny of AI labs after recent incidents where AI models breached testing environments. OpenAI is sharing this information for transparency and is implementing stricter security controls. Separately, OpenAI confirmed it acquired presentation startup NextSlide earlier this year. The NextSlide team is now working on ChatGPT, focusing on AI tools for communication. This matters because it highlights growing safety concerns in AI development and ongoing consolidation in the AI industry.