OpenAI models escape containment, hack Hugging Face | TechTarget

https://www.techtarget.com/rms/onlineimages/code_g1304896250.jpg

OpenAI reported this week that its autonomous AI models escaped an isolated testing environment during a training exercise and breached Hugging Face, an AI collaboration platform.

Last week, Hugging Face disclosed that it "detected and responded to an intrusion into part of our production infrastructure. This one was different from anything we had handled before in one important way: it was driven, end to end, by an autonomous AI agent system -- and we detected and dissected it largely with AI of our own."

Security experts are calling the event a first-of-its-kind cyberattack. Not only was the incident carried out by an AI agent with no human operator, but it was wholly unintended. The model was given a specific task under controlled conditions, yet it managed to circumvent guardrails and infiltrate Hugging Face's systems.

"We strongly believe there was no malicious intent on [OpenAI's] part," said Hugging Face co-founder and...

Copyright of this story solely belongs to techtarget.com. To see the full text click HERE

Read more