WASHINGTON: Geoffrey Hinton’s AI warning has renewed attention on regulation after the Nobel laureate told U.S. lawmakers they may have about a year to establish stronger safeguards.
Hinton spoke Wednesday after a closed-door Capitol Hill briefing with lawmakers and other AI experts. He said predictions for the arrival of highly capable “superintelligence” have shortened significantly in recent years.
However, Hinton’s timeframe is an expert assessment, not a demonstrated deadline. Researchers continue to disagree over when, or whether, systems surpassing human capabilities across many areas will emerge.
Hinton said AI systems have reached a stage where they can help design improved AI systems. He described that process as “recursive self-improvement.”
As a result, Hinton argued that lawmakers should act before increasingly capable systems become harder to control. He also called for slowing development while stronger safeguards are created.
Sen. Bernie Sanders of Vermont organised the private briefing and invited lawmakers from both parties. Previously, Hinton, physicist Max Tegmark and researcher Ajeya Cotra would brief senators on September 16.
Read: Dario Amodei AI Slowdown Plan Calls for External Oversight
Meanwhile, lawmakers discussed recent AI security incidents. Rep. Ted Lieu of California cited an OpenAI-related Hugging Face incident while arguing for stronger emergency controls.
Separately, Lieu and Republican Rep. Nathaniel Moran of Texas introduced the AI Kill Switch Act in July. The proposal would require covered developers to maintain the ability to throttle, suspend or shut down powerful AI systems.
The bill would also establish a government process for ordering emergency restrictions when an AI system poses a risk of catastrophic harm. However, the proposal has not become law.
At the same time, Sen. John Kennedy of Louisiana pursued a separate Senate proposal requiring human-controlled shutdown mechanisms in AI systems.
Senate records show Kennedy sought unanimous consent on September 16, while Sen. Rand Paul of Kentucky objected to his request.
Read: Jacob Coxon Resigns, Says AI 10% Chance of ‘Killing All Humans’
Still, broader federal AI legislation remains unsettled. Senate lawmakers have continued discussions over catastrophic-risk safeguards.
The Geoffrey Hinton AI warning comes amid renewed debate after several AI developers disclosed unexpected model behaviour during safety testing.
Yet researchers and policymakers continue to differ over which risks require regulation and how quickly Congress should respond.
