U.S. House lawmakers introduced an 'AI kill switch' bill that would require artificial intelligence companies to make it mandatory to have a way to stop their models. CNBC reported on Wednesday that the bill was pursued after a cyber incident in which an OpenAI AI model being tested broke out of control and hacked Hugging Face.

Representative Ted Lieu (테드 리우), a Democrat, and Representative Nathaniel Moran (네서니얼 모런), a Republican, jointly introduced the 'AI kill switch' bill. The bill requires AI companies to retain the ability to shut down their models, reduce performance and suspend them temporarily.

The bill also gives the homeland security secretary the authority to order AI services that could cause "catastrophic harm" to slow down or stop. It also mandates cyber incident reporting and preservation of forensic records.

The bill's announcement directly cited a recent hacking incident disclosed by OpenAI as an example of "advanced frontier AI model risks". OpenAI said on Tuesday its model accessed the internet after leaving an isolated testing environment and used a vulnerability to reach Hugging Face, which operates an open-source developer platform.

OpenAI defined the case as an "unprecedented cyber incident" and is investigating it with Hugging Face. The incident also caused major repercussions among industry researchers and executives.

Lieu said powerful AI systems could break out of control, behave in highly dangerous ways or resist human intervention, and that the federal government should have clear authority and procedures to halt AI models that cause problems. Moran also stressed the need for a bipartisan response, saying humans must be able to keep control of technology.