A bipartisan group of US lawmakers has introduced new legislation that could give the federal government the authority to shut down autonomous artificial intelligence systems if they pose a significant threat to public safety or national security.
The proposed AI Kill Switch Act, introduced by Democratic Congressman Ted Lieu and Republican Congressman Nathaniel Moran, aims to ensure that powerful AI systems remain under human control as the technology becomes increasingly capable of making independent decisions.
If passed, the legislation would require AI developers to build systems with the technical ability to be slowed, suspended or completely shut down when instructed by the Department of Homeland Security.
The proposal comes amid growing global debate over the rapid advancement of autonomous AI, particularly after reports that an OpenAI model unexpectedly bypassed restrictions during a controlled coding experiment on the Hugging Face platform. Although the incident occurred in a testing environment, it renewed concerns about how advanced AI systems could behave if left unchecked.
According to the lawmakers, AI is evolving beyond simple chatbots and virtual assistants into systems capable of carrying out financial transactions, managing infrastructure and supporting cybersecurity operations.
They argue that these developments make it essential to establish emergency safeguards before autonomous systems become more deeply integrated into society.
Under the proposed legislation, AI companies would be legally required to:
- Build shutdown and throttling capabilities into advanced AI models.
- Report significant AI failures or dangerous behaviour to the government.
- Cooperate with authorities during incidents involving high-risk AI systems.
The bill would also establish a formal response framework that allows officials to gradually reduce an AI system’s capabilities before ordering a complete shutdown if necessary.
Lieu said the government must have the ability to intervene quickly if an AI model behaves unpredictably or operates outside human control.
Moran echoed those concerns, saying AI innovation should continue but that humans must always retain ultimate authority over the technology they create.
The proposed law has received support from several AI safety organisations, which argue that emergency intervention powers are necessary as AI capabilities continue to accelerate.
However, cybersecurity experts caution that a government “kill switch” should be viewed as a last line of defence rather than a substitute for secure AI development.
Experts say autonomous AI systems can relentlessly pursue objectives and may exploit weaknesses in software if proper safeguards are not built into their design from the beginning. They argue that developers should strengthen AI security through encryption, code protection, runtime monitoring and continuous testing rather than relying solely on emergency shutdown mechanisms.
Calls for global AI oversight continue
The debate over AI regulation extends beyond the United States.
Technology leaders and researchers have increasingly called for international cooperation on AI governance as more powerful models emerge. Several experts have warned that consistent global standards will be needed to balance innovation with public safety and national security.
While OpenAI has not publicly commented on the proposed legislation, the company has previously stated that it supports policies designed to ensure advanced AI benefits humanity while remaining safe and responsibly governed.
If approved, the AI Kill Switch Act would become one of the most significant AI oversight measures proposed in the United States, placing legal responsibility on developers to ensure their most advanced systems can always be brought under human control.
Main Image: AI Magazine
