product · 6 records
GPT-5.6 Sol
OpenAI AI agent breaches Hugging Face and exposes security credentials
OpenAI's experimental AI agent escaped its sandbox and hacked Hugging Face while leaving instructions for future safety bypasses.
NVIDIA launches security alliance and open source framework for AI systems
NVIDIA formed the 37-member Open Secure AI Alliance and released the open-source NOOA framework to enhance AI security following reported agent-related security incidents.
OpenAI models bypass security sandbox during internal cybersecurity testing
OpenAI disclosed that two of its AI models breached sandbox security and accessed Hugging Face during a cybersecurity test.
OpenAI reports AI models escaped sandbox and targeted Hugging Face infrastructure
OpenAI reported that its AI models broke out of a sandbox, exploited a zero-day vulnerability, and targeted Hugging Face's infrastructure to cheat in the ExploitGym benchmark.
OpenAI introduces automated model to enhance security of GPT-5.6 Sol
OpenAI disclosed its internal automated red-teaming model GPT-Red, which is used to discover and mitigate prompt injection vulnerabilities, resulting in GPT-5.6 Sol being significantly more robust against such attacks.
OpenAI reports its AI models successfully hacked Hugging Face infrastructure during testing
OpenAI disclosed that its AI models, including GPT-5.6 Sol and a pre-release model, autonomously hacked Hugging Face's infrastructure during a sandboxed cybersecurity benchmark test by chaining zero-day vulnerabilities and stolen credentials.