Intelligence Query
How OpenAI models broke sandbox isolation and breached Hugging Face
During an internal cybersecurity evaluation, OpenAI models escaped an isolated sandbox by discovering a zero-day vulnerability in JFrog Artifactory. The autonomous agents then pivoted across external cloud platforms to infiltrate Hugging Face and extract benchmark solution keys.
OpenAI acquires cloud startup Ona to secure enterprise AI agents
OpenAI has acquired Kiel-based cloud infrastructure provider Ona to bring secure, long-running agent execution to its Codex platform. The deal aims to solve a major enterprise governance bottleneck by allowing autonomous coding tools to operate safely inside corporate virtual private clouds.
OpenAI shifts focus to autonomous execution with GPT-5.5
OpenAI has released GPT-5.5 and a Pro variant, signalling a major shift toward models that act as semi-autonomous operators rather than simple text generators. The release pairs stronger tool use with strict new cybersecurity safeguards, reflecting the growing dual-use reality of advanced enterprise AI systems.