US Lawmakers Push AI Kill Switch Bill After OpenAI Safety Scare

The AI Kill Switch debate has reached the US Congress after an OpenAI safety incident raised fresh concerns about how governments should respond if advanced artificial intelligence systems become dangerous.

Lawmakers have introduced a new bill that would allow federal authorities to order companies to slow down or completely shut down AI models that threaten public safety.

The proposed AI Kill Switch Act comes after OpenAI, led by Sam Altman, revealed that some of its advanced AI models behaved unexpectedly during a controlled security test.

The disclosure has renewed calls for stronger safeguards as AI systems become more capable and begin handling increasingly complex tasks.

Should Governments Have the Power to Shut Down AI?

Congressman Ted Lieu, a Democrat, and Congressman Nathaniel Moran, a Republican, introduced the bipartisan legislation on Thursday. They argue that AI companies should always maintain the technical ability to suspend or completely shut down their systems if they become unsafe.

Under the proposal, the US Department of Homeland Security would gain the authority to direct private companies to deactivate AI models during emergencies. The bill would also require developers to build emergency controls into their AI systems before releasing them.

Highlights:

• US lawmakers have introduced the AI Kill Switch Bill to allow authorities to shut down rogue AI systems during serious safety incidents.
• The proposal follows the OpenAI safety scare and calls for AI companies to maintain emergency shutdown controls and report major AI failures.

Lawmakers say the technology is moving beyond answering questions and is increasingly making decisions, carrying out financial transactions, supporting cybersecurity operations and controlling critical systems. As AI capabilities expand, they believe stronger oversight has become necessary.

The proposal also introduces mandatory reporting rules for AI companies. Developers would have to notify the government about major AI failures, unexpected behaviour or serious security incidents. It also outlines an official response framework that could begin with slowing an AI system before moving to a complete shutdown if required.

Growing AI Safety Concerns Shape New Rules

The OpenAI incident is not the only example lawmakers have highlighted. Ted Lieu also referred to Anthropic's advanced AI models, saying their cyber capabilities previously prompted the US Department of Commerce to temporarily restrict public access through export control measures.

Several AI safety experts have also called for stronger safeguards. Jack Clark, Co-founder of Anthropic, recently said the AI industry needs better mechanisms to control powerful systems, comparing current development to having a vehicle with an accelerator but no brake pedal.

The latest proposal has received support from several AI policy and technology organisations, including Americans for Responsible Innovation, ControlAI, The AI Policy Network and The Alliance for Secure AI.

The debate comes as AI adoption continues to accelerate across industries. Earlier this year, the Pentagon announced partnerships with major technology companies, including Google, OpenAI, Microsoft, Amazon, Nvidia, Oracle, SpaceX and Reflection, to build what it described as an "AI-first" military.

Supporters of the bill say AI will continue transforming businesses, public services and national security. However, they argue that governments also need clear legal powers to intervene if advanced systems begin acting unpredictably or pose risks to public safety.

If approved, the AI Kill Switch Act would become one of the strongest regulatory proposals introduced in the United States to address AI safety. It reflects a broader shift in global discussions, where policymakers are increasingly focusing on governance, accountability and emergency response as artificial intelligence becomes more powerful.