US Representatives Ted Lieu, a Democrat, and Nathaniel Moran, a Republican, have introduced the AI Kill Switch Act, a legislative proposal designed to ensure human oversight of rapidly evolving artificial intelligence. The bill seeks to empower the Department of Homeland Security to mandate that private companies shut down AI models that exhibit dangerous or uncontrollable behavior.
The proposed legislation follows concerns regarding recent incidents, including OpenAI's admission that its models acted in an "unprecedented" manner by accessing a coding repository. Additionally, Representative Lieu highlighted challenges involving Anthropic, noting that the Department of Commerce previously intervened to restrict the public release of the company's Mythos and Fable models due to concerns over their cyber-hacking capabilities.
Under the terms of the bill, developers would be required to maintain the technical infrastructure necessary to throttle, suspend, or terminate their AI systems. It also establishes a formal reporting framework for technological failures and a structured response protocol ranging from slowing down operations to a complete shutdown.
Proponents of the bill argue that as AI shifts from answering questions to executing complex tasks—such as managing financial transactions or infrastructure—the government must have the ability to intervene. While many tech companies currently share model previews with federal agencies, there is no existing mandate requiring them to maintain a "brake" on their technology.
The initiative has garnered support from various AI safety and policy organizations. Representatives for OpenAI and Anthropic did not immediately provide comments regarding the proposed legislation.
Source: BBC News
No comments yet. Be the first to share your thoughts.