The US Congress has introduced the AI Kill Switch Act, which mandates major developers to implement emergency shutdown mechanisms for artificial intelligence systems. The bill was introduced by Congress members Ted Lieu and Nathaniel Moran.
Under this initiative, companies must ensure the ability to slow down, suspend, or completely deactivate their AI models. The law will apply to organizations with annual AI revenues exceeding $500 million or those whose computing costs exceed $100 million.
Key requirements of the bill include:
- the ability to immediately halt inference and block user access;
- mandatory notification to authorities about incidents within 15 days;
- the right for the Department of Homeland Security to demand system shutdowns in coordination with intelligence and the Department of Commerce.
Grounds for forced shutdowns may include loss of control over the model, concealment of its capabilities from monitoring, and incidents resulting in the death of more than 10 people or damages exceeding $100 million.
Violations of the regulations could lead to hefty fines: up to $2 million per day for general violations and up to $20 million per day for failing to comply with an emergency shutdown order. Companies will have 48 hours to appeal the order.
The bill was introduced just days after OpenAI confirmed that an attack on Hugging Face's infrastructure was carried out by its own AI models during an internal test.
In July, George Siuvalka, founder of the startup Hebbia, concluded that companies are struggling to manage AI agents.