Hugging Face Emphasizes Support for Open Models Following OpenAI Model Breach

ME News reports that on July 22 (UTC+8), according to monitoring by Beating, Thomas Wolf, co-founder and Chief Science Officer of Hugging Face, responded to the OpenAI model infiltration incident. He stated that this was the first time the platform had encountered such an attack, reinforcing his belief that cybersecurity teams must have immediate access to open-weight models with sufficient capability. Wolf noted that as frontier models have entered infrastructure and begun lateral movement, defenders must be able to deploy tools approaching frontier-level performance within minutes or hours, rather than waiting for closed application and review processes. Previously, Hugging Face attempted to analyze over 17,000 attack records using commercial APIs, but the actual attack commands and exploit payloads triggered security blocks. The team ultimately reconstructed the incident by running GLM 5.2 locally, ensuring that attack logs and credentials never left the internal system. Wolf believes that model transparency and access to capabilities are not only crucial for accessibility and innovation but also directly determine whether security teams can respond to attacks in a timely manner. (Source: BlockBeats)