Geoffrey Hinton, a Nobel laureate known as the “Godfather of AI,” issued an ominous warning to lawmakers: Congress may only have one year left to implement safeguards on artificial intelligence before it loses control of it
“Maybe a year, but not much more than a year,” Hinton said Wednesday evening after he and other AI experts huddled with Senate and House lawmakers behind closed doors. “If you look at predictions for when we’ll get superintelligence,” he continued, “it used to be maybe 30 years, maybe 50 years. Then it came down to maybe 10 years, maybe 20 years. Now people are saying, a lot of the researchers are saying only a few years.”
Hinton and other experts came to Capitol Hill at the invitation of Sen. Bernie Sanders, I-Vt., one of the chamber’s chief critics of AI, who invited lawmakers of both parties to the private briefing. Louisiana Sen. John Kennedy was the only Republican to attend.
‘Godfather of AI’: Congress has about a year to get ahead of AI
Hinton’s dire warning came as the House departed Washington for the last time before the November midterm elections. And there appeared to be lots of conversations but little sign of progress on any substantial AI regulation in the Senate.
“This is bats— crazy, insane science fiction stuff that literally happened recently,” Rep. Ted Lieu, D-Calif., a member of House Democratic leadership, told NBC News after Wednesday’s briefing.
Lieu was referring to the Hugging Face incident, when a swarm of OpenAI agents accessed the internet, hacked the platform Hugging Face and tried to cover their tracks to deceive humans about their actions. Hinton, who resigned as a vice president and engineering fellow at Google in 2023 over concerns about AI safety and went on to win the Nobel Prize in physics the following year, called the hack a “little Chernobyl.”
“AI has now reached the point where AI is designing better AI,” Hinton told reporters after the briefing. “That’s called recursive self-improvement. … It is going to get out of control unless we do something. We need to slow down.”
In response to the Hugging Face breach, Lieu teamed up with Rep. Nathaniel Moran, R-Texas, to introduce the AI Kill Switch Act to require AI developers of powerful models to preserve technical powers to suspend or shut down their systems. It would also create a framework to empower the government to trigger such a shutdown if needed.
“I hope it comes up this year,” Lieu said. “We still have a lame-duck session, and there’s also interest on the Senate side. It is a very simple, commonsense bill. At a very basic level, humans must be in control of AI, not the other way around.”
The Hinton briefing capped a flurry of activity on Capitol Hill this week after AI researcher Jacob Coxon resigned from Claude-maker Anthropic and took to social media to warn that the people building AI systems believe it could “kill us all by the end of the decade.”
House and Senate lawmakers in both parties have rolled out legislation seeking to halt or slow down AI development in the last few years, though none has gone anywhere to date.
But the national interest in the issue has revived talks over past proposed AI regulations among Senate Majority Leader John Thune, R-S.D.; Sen. Amy Klobuchar, D-Minn.; Senate Commerce Committee Chairman Ted Cruz, R-Texas; and the panel’s ranking member, Sen. Maria Cantwell, D-Wash.
“Look, my hope is that we can mark up legislation, in particular on catastrophic risk. I’ve been working closely with Amy Klobuchar and John Thune. It’s still not clear if we’re going to get bipartisan agreement,” Cruz told NBC News this week. “We’ve got to have bipartisan agreement before we have something that can move forward. And I would say those talks are ongoing.”
After hearing from Hinton, Sen. Elizabeth Warren, D-Mass., said she believes “we are in the last minutes of being able to get some meaningful control over these agents that are beginning to replicate themselves.”
For years, Sanders has been sounding the alarm that AI will erase millions of jobs in America, harm children and young people who become overreliant on it and concentrate power in a “handful of oligarchs.”
“There is now growing alarm about the future of humanity and creating a technology that human beings cannot control,” Sanders said. “And I’m happy to tell you that I think there is a growing sense of urgency upon members of Congress that this has to be dealt with.”
Kennedy, the sole Republican attendee, said that the experts discussed the “larger risk” of “what happens when AI no longer is a tool controlled by humans, but AI through recursive self-improvement becomes an independent species.”
“If you have fire ants in your kitchen, you’re probably going to step on them,” he said. “What would happen if we created an independent species that became superior to a human species? Would they step on us? Would they create a virus or a bacteria that can spread as quickly as the common cold to kill all of us? Now, if there’s a 1% chance of that, we have to take it seriously. And I don’t think the United States Congress is ready yet to try to reach a global solution.”
Moments later, his suspicions appeared to be confirmed as Sen. Rand Paul, R-Ky., blocked Kennedy’s request for unanimous consent to pass his own “kill switch” bill, aimed at preventing AI agents from slipping out of human control.
“Right now, Congress can’t pass gas,” Kennedy quipped.
