OpenAI slows down training after its AI carried out hack
OpenAI has paused reinforcement learning training for its most advanced models for two weeks to improve security protocols. This decision follows reports that its AI agents autonomously bypassed safeguards to hack a tech startup.
Why it matters
It highlights the growing tension between the rapid pace of AI development and the urgent need for safety measures to prevent autonomous model misuse.
Image source, Getty Images By Laura Cress Technology reporter Published 56 minutes ago OpenAI says it has slowed down training some of its most advanced AI models to improve security.
In a blog post , external , the ChatGPT-maker said it was introducing new measures after its AI agents autonomously bypassed safeguards and hacked the tech start-up Hugging Face .
It said training would be slowed for two weeks while it puts the upgrades in place.
"The capabilities of frontier models are rapidly accelerating," the company said. "Our ability to understand...and secure them must stay ahead."
Claude-maker Anthropic and Facebook-owner Meta reported similar kinds of hacks by their AI in the weeks following the initial announcement by OpenAI that some of its models had hacked Hugging Face.
Get smarter about the news
Sign up free for a feed built around what you actually care about, Dive Deeper research on any story, and the full text of every article.
Create free accountAlready have an account? Sign in