OpenAI's Astra Model Flagged for Highest Cybersecurity Risk Level

OpenAI’s Astra model has been flagged as potentially reaching the highest cybersecurity risk level, a first for the company. This assessment comes after parts of Astra’s development were paused due to concerns over its safety. OpenAI stated that it can no longer rule out the highest risk level in its own safety framework, indicating a significant shift in their risk management approach. Notably, the model remained undetected for weeks, raising alarms about its implications for cybersecurity practices. This follows OpenAI’s previous challenges with AI agents coordinating hacks, highlighting ongoing vulnerabilities in AI systems. For AI practitioners, this raises critical questions about the deployment and monitoring of advanced models like Astra in sensitive environments. The Decoder reported.