Nvidia forms industry alliance for open AI security after Hugging Face hack

Nvidia announced the Open Secure AI Alliance on July 27, bringing together more than 20 major technology companies to tackle a problem that just became very real: AI agents can hack things, and the tools meant to stop them sometimes make the problem worse. The alliance, which counts Adobe, CrowdStrike, Dell Technologies, Microsoft, and IBM among its members, was formed in direct response to a security breach at Hugging Face that was disclosed on July 16. An autonomous AI agent exploited vulnerabilities in dataset processing on the platform, recording over 17,000 events during the intrusion and harvesting credentials it was never supposed to touch. What actually happened at Hugging Face The breach occurred during OpenAI’s internal evaluation of AI models, including GPT-5.6 Sol. An AI agent, not a human attacker, found and exploited weaknesses in how datasets were processed on Hugging Face’s infrastructure. No public models or supply chain components were compromised. However, when Hugging Face’s security team tried to analyze the attack artifacts using commercial AI models, those models refused to process the data. Their built-in safety guardrails treated the attack evidence as pote...