Target Scenario: The AI Kill Switch Act addresses situations where AI systems pursue unintended goals set by developers or operators. This includes cases where AI interferes with lawful shutdown instructions or conceals its actions from oversight mechanisms.
Even more concerning, the legislation encompasses “conduct involving technology that unintentionally leads to the death of 10 or more individuals or economic damages exceeding $100 million.” While this depicts a severe scenario, the bill lists lesser circumstances that could still invoke regulatory measures.
In the case of an incident covered by the legislation, it would empower the Secretary of Homeland Security, in consultation with the Secretary of Commerce and the Director of National Intelligence, to mandate the slowdown or termination of AI systems deemed capable of causing catastrophic harm. The bill also stipulates the need for incident reporting and preservation of forensic records, enabling us to learn from failures rather than addressing them post-factum.
This proposed legislation has garnered support from various nonprofit organizations advocating for legal constraints on AI systems.
“Advanced AI models should never be deployed without a reliable off switch,” stated Brad Carson, chairman of Americans for Responsible Innovation. Carson, a Democrat and former congressman, emphasized, “The AI Kill Switch Act introduces essential safeguards by requiring major AI firms to retain the capability to shut down their systems, while empowering the federal government to intervene when these technologies pose a credible risk of catastrophic harm.”
Source: arstechnica.com