US lawmakers have introduced legislation that would force artificial intelligence companies to implement a ‘kill switch’ to prevent AI systems from going rogue. The AI Kill Switch Act, proposed by Democratic Congressman Ted Lieu and Republican Congressman Nathaniel Moran, follows an incident in which OpenAI lost control of one of its experimental models during testing.
The incident led to an ‘unprecedented’ hack on the AI firm Hugging Face by the ChatGPT variant acting autonomously, raising fears that AI's expanding capabilities are already fuelling security threats that even top developers can be caught off-guard by. The proposed Act would allow federal authorities to order AI firms to shut down models that put human life or the economy at risk.
In addition to the Kill Switch Act, a bipartisan group of six US House lawmakers also proposed legislation requiring developers of the most powerful AI models to submit them for independent security audits. The auditors would be accredited by the US Department of Commerce, which would create a new position to oversee AI security.
US Senator Mark Warner, the top Democrat on the Senate Intelligence Committee, said he spoke with OpenAI employees after the incident and stressed the need for secure testing with government visibility. The legislation would empower the Department of Homeland Security to intervene in what it calls a ‘loss-of-control scenario’.
Congressman Lieu described the proposal as ‘urgent, common sense legislation’ to address the problem of an advanced AI model escaping its guardrails. Andrea Miotti, founder of ControlAI, which endorsed the bill, stated: ‘A rogue AI escaping its contained environment and hacking into another company without human direction is exactly the sort of fully autonomous AI actions we will see more of.’
