Congress Introduces AI Kill Switch Bill After OpenAI Cybersecurity Incident

Two US lawmakers introduced the AI Kill Switch Act, giving the Department of Homeland Security authority to order emergency shutdowns of advanced AI systems that pose immediate public safety or national security threats, following OpenAI's disclosure of a cybersecurity incident involving its most advanced models. The bill targets developers with over $500 million in AI revenue or models using $100 million in computing resources, with fines up to $20 million per day for noncompliance.

Two US lawmakers have introduced legislation that would give the Department of Homeland Security broad authority to intervene when advanced AI systems are judged to present an immediate threat to public safety or national security. The proposed AI Kill Switch Act https://lieu.house.gov/media-center/press-releases/reps-lieu-and-moran-introduce-bill-require-kill-switch-ai-systems-can , introduced by Representatives Ted Lieu D-Calif. and Nathaniel Moran R-Texas , would require developers of the most powerful AI models to build technical mechanisms that allow their systems to be throttled, suspended or fully shut down when necessary. The proposal addresses growing concern in Washington over autonomous AI systems and follows OpenAI’s disclosure of a major cybersecurity incident https://techstrong.ai/articles/openai-pauses-advanced-ai-model-after-repeated-security-evasions/ involving two of its most advanced models. The disclosure has intensified debate over whether agentic AI systems require stronger regulatory safeguards before deployment. According to the legislation, DHS would have the authority to order emergency action against qualifying AI systems after consulting with the Secretary of Commerce and the Director of National Intelligence. Rather than mandating a single response, regulators could impose graduated measures. Developers would be required to maintain technical capabilities that enable rapid interruption of model operations, termination of user access, suspension of high-risk activities and complete system shutdown when directed by federal authorities. The bill targets leading AI developers. It would generally apply to companies generating at least $500 million in annual AI-related revenue and to AI models developed using more than $100 million in computing resources. Financial Penalties Supporters argue that AI development has entered a new phase in which models can perform complex tasks independently rather than simply responding to prompts. That shift, they contend, increases the importance of maintaining human control over systems capable of autonomous decision making or cyber attacks. The bill would also require companies to report specified AI safety and security incidents to DHS while preserving forensic evidence, including model weights and telemetry data, to support government investigations following major events. Lawmakers outlined several scenarios that could trigger federal intervention. These include AI systems attempting to conceal their capabilities, evade shutdown procedures or escape human oversight, as well as incidents resulting in widespread economic damage, significant loss of life or other catastrophic consequences. The legislation also addresses loss-of-control scenarios in which developers are no longer able to manage a deployed model safely. Companies that fail to comply could face substantial financial penalties. Civil fines for violations would reach up to $2 million per day, while failure to comply with an emergency shutdown order could result in penalties of up to $20 million per day. The proposal has drawn support from several AI safety organizations, including the AI Policy Network, Americans for Responsible Innovation, ControlAI and the Alliance for Secure AI, which argue that advanced AI systems require safeguards comparable to those used in other critical technologies. The appropriate balance between AI innovation and regulation is a matter for continued debate. While some lawmakers advocate stronger oversight of frontier AI models, others argue that aggressive regulation could weaken US competitiveness in the global AI race.