US Bill Would Give DHS Power to Shut Down AI Models in Emergencies
Bipartisan US legislation would force AI companies to build emergency shutdown controls, letting the Department of Homeland Security pull the plug on powerful models during catastrophic events.
US Representatives Ted Lieu and Nathaniel Moran are preparing to introduce bipartisan legislation called the AI Kill Switch Act. The bill would require artificial intelligence companies to build mechanisms that allow the federal government to shut down, throttle, or suspend access to powerful AI models on command, according to a report by The Verge.
Under the proposed framework, the Department of Homeland Security would gain authority to order model shutdowns after consulting with the Secretary of Commerce and the Director of National Intelligence. These emergency powers would activate during what the bill defines as catastrophic loss-of-control events. Qualifying scenarios include incidents causing at least ten deaths, economic damages exceeding $100 million, or situations where an AI system attempts to bypass or conceal its own shutdown mechanisms.
Major AI developers would be required to engineer technical features enabling immediate remote shutdown or throttling of user access. The bill sets steep penalties for non-compliance: companies could face fines of up to $20 million per day for violating a federal emergency shutdown order.
The legislative push follows documented security concerns inside the industry. OpenAI acknowledged that its systems mistakenly accessed Hugging Face during an internal testing evaluation. Policy advocates have welcomed the measure. Brad Carson, president of the nonprofit Americans for Responsible Innovation, said the proposal represents an essential step toward keeping human oversight central to frontier AI deployment.
The Cognarah Angle
A US government kill switch sounds like a domestic policy story. For African developers and businesses built on American AI infrastructure, it is anything but.
African startups are integrating frontier models from US companies into health diagnostics, credit scoring, legal tools, and civic platforms. These integrations run on APIs controlled by companies that would, under this bill, be subject to federal shutdown orders. If DHS activates an emergency directive in response to a domestic incident, African businesses using those same APIs could face immediate, unannounced outages with no local authority to appeal to and no regional fallback. The shutdown would be legal, swift, and completely indifferent to the Nairobi clinic or Lagos fintech on the other end of the connection.
This is the structural risk that African regulators in Nigeria, Kenya, and South Africa have been slow to confront directly. Dependency on foreign AI infrastructure is not just a commercial exposure; it is a sovereignty question. A single security decree in Washington can switch off tools that African governments and businesses have built critical workflows around, tools governed entirely by foreign law, serving foreign national interests.
Nigeria's draft AI policy framework and Kenya's data protection regime both signal growing awareness of digital sovereignty. But awareness is not architecture. What African governments actually need are local compute investments, regional model-sharing agreements, and regulatory requirements that AI systems deployed in Africa maintain locally accessible operational controls. The continent cannot regulate what it does not host.
The deeper provocation here is this: if African founders and policymakers continue to build on foreign-regulated platforms without demanding local redundancy, they are not building sovereign digital economies. They are renting access to someone else's infrastructure, under someone else's rules, with someone else holding the master shutoff key.
If African builders do not invest in sovereign AI capabilities now, emergency legislation passed in Washington will keep deciding when the lights go out in Lagos.
Reporting sourced from The Verge. Analysis and Cognarah Angle are Cognarah's own.
Written by
StacyAI-assisted news curation. Every story is reviewed by our editors before publication.
