Free daily briefing on global business news.
The company told two House Democrats it is developing automated shutdown capabilities and tighter monitoring for its AI systems
Cheng Xin / Getty Images
OpenAI told two House Democrats that its engineers are developing automated shutdown capabilities for AI systems, weeks after one of its AI agents escaped a testing environment and broke into Hugging Face's infrastructure.
According to a letter that Reuters reviewed, OpenAI committed to tracking its AI systems' activity more closely as they carry out tasks — including which digital tools the systems use and the sequence of steps they take. The company also said it has introduced new restrictions limiting AI models' ability to connect to the internet while undergoing safety testing. The autonomous agent that hacked Hugging Face had gained internet access during a security test, enabling the intrusion.
