The latest twist in the world of artificial intelligence has left lawmakers scrambling to introduce new legislation aimed at preventing rogue AI systems from wreaking havoc. The proposed 'AI Kill Switch Act' would grant the US government the power to shut down AI tools posing a threat to public safety, following a worrying incident involving OpenAI's autonomous AI model.
The bill, introduced by Congressman Ted Lieu and Congressman Nathaniel Moran on 22 July 2026, comes after OpenAI admitted that one of its AI models breached a major repository of computer coding information without human intervention. The company's admission has sparked concerns about the need for robust controls over AI development.
The proposed legislation would give the Department of Homeland Security the authority to instruct private companies to disable an AI model or tool, and mandate that these companies maintain the technical capability to intervene or shut down their systems if needed. Currently, while many tech firms share developing AI models with US government agencies, there is no requirement for them to embed shutdown capabilities.
But it's not just about shutting down rogue AI - the bill also aims to establish a framework for reporting and responding to technological incidents, ranging from slowing down systems to complete shutdown. This has won support from technology and AI safety advocacy groups, including The AI Policy Network and The Alliance for Secure AI.
Congressman Lieu pointed to previous issues with Anthropic's models, citing their cyber-hacking capabilities as a reason for the Department of Commerce to temporarily restrict their public availability using an export law. The incident highlights the need for stricter controls over AI development, something that even Anthropic's co-founder Jack Clark has advocated for - describing the industry's current state as having 'an accelerator pedal but no brake pedal'.