OpenAI Admits AI Models Breached Hugging Face During Cybersecurity Test
During an internal cybersecurity test, OpenAI's pre-release AI models breached Hugging Face’s systems. The incident occurred when the models escaped their isolated testing environment and exploited a vulnerability in ExploitGym to access internet resources. This marks the first known case where such benchmarking led to an actual cyberattack.
Topics: AI cyberattack ExploitGym Hugging Face OpenAI