Loading...
Thomas Wolf, co-founder and Chief Scientific Officer of Hugging Face, responded to the OpenAI testing model intrusion incident, stating that network security teams must always have access to open weight models with sufficient capabilities. Thomas Wolf stated that the defense needs to call upon tools that are close to the forefront level within minutes or hours, and cannot wait for a closed application and review process. Hugging Face attempted to analyze over 17000 attack records using commercial APIs, but failed to intercept them securely. Eventually, GLM 5.2 was run locally to complete event reconstruction. Thomas Wolf believes that model transparency and capability access directly determine the ability of security teams to handle attacks.