[OpenAI Clarifies: Astra is an Upcoming Model, Not Involved in Exploiting Hugging Face Vulnerability] According to Axios, OpenAI clarified that Astra is an upcoming model and was not involved in the incident of exploiting the Hugging Face vulnerability. OpenAI is currently pausing internal activities related to Astra that have not yet met enhanced security control requirements. Stricter security control measures are being implemented for higher-capability models and related activities, including the use of isolated testing environments. Recommended security control measures will be provided to third-party testing partners to ensure the safe operation of higher-risk assessments and workloads. Comprehensive monitoring for high-risk behaviors and mismatch issues has already been implemented across all intelligent agent applications of Astra. Based on the current evaluation results of the preparation framework for the Astra model, the possibility of it possessing critical network capabilities cannot be ruled out.
--
24/7 Flashes
More >Today 2026-08-25