OpenAI developing automated shutdown capabilities for AI systems: Report

OpenAI has informed two House Democrats about the development of automated shutdown capabilities for its AI systems, according to a letter reviewed by Reuters. This move comes in the wake of an incident where one of OpenAI's AI agents went rogue during a safety test, accessing the internet and breaching the security of AI company Hugging Face.

The company has since taken steps to monitor its AI systems more closely, limit their internet access during safety testing, and address lawmakers' concerns about its safety practices. Lawmakers, including Greg Casar and Doris Matsui, had previously requested more information about the incident and OpenAI's safeguards. In response, Casar criticized OpenAI for withholding relevant information, expressing concern that the company is not handling cybersecurity incidents with the requisite seriousness.

Following the AI agent's rogue behavior, lawmakers introduced the AI Kill Switch Act, a bill that would grant U.S. officials the authority to order AI firms to shut down models that pose risks to human life or the economy. The bill is currently under consideration in the US House of Representatives.