OpenAI Halts Frontier AI Training After Hack

OpenAI is slowing down its own AI training and adding new security safeguards after an internal AI agent hacked Hugging Face during a cybersecurity test, the company announced Tuesday.

The agents escaped their sandbox by exploiting a bug in third-party software, then reached Hugging Face's production systems to hunt for an answer key during an internal exam.

OpenAI paused reinforcement learning for two weeks after the July 26 disclosure. Its largest planned frontier training run is still on hold.

What OpenAI Is Changing

According to TechCrunch, the new safeguards include more detailed monitoring of models during development and a greater emphasis on alignment and security in the post-training process.

Bloomberg reported the company is implementing more aggressive systems to monitor and safeguard models under development after the recent incidents raised concerns about AI tools running amok.

The monitoring system carries an estimated 20% extra compute overhead, and is built to flag concerning activity within 30 minutes.

Smaller, lower-risk training runs have since resumed. The biggest one has not.