Research4h ago
Research
OpenAI Releases Astra Cybersecurity Evaluations
Preliminary evaluations for Astra detail the model's current cybersecurity capabilities and the specific safeguards OpenAI is implementing. The report outlines security controls designed to prevent misuse in cyberattacks. This transparency provides a baseline for practitioners assessing the risk profile of multimodal models. It remains a cautious, incremental update on safety protocols.