A string of new, terrifying warnings from the artificial intelligence community has put Congress, the White House and the A.I. industry on notice: Slow the progress of the rapidly developing technology before a swarm of bots takes over the Internet and “kill all humans.”
President Trump and House Speaker Mike Johnson responded to the shocking AI threats on Sunday, acknowledging that “guardrails” on AI are needed, or in the president’s case, acceptable.
“We can put up guardrails. We can do this and that,” Mr. Trump said while visiting Ireland on Sunday. “But I think you have a lot of negative forces that are bringing up things that won’t happen.”
Mr. Trump tepidly agreed to federal regulations on AI just hours after Anthropic CEO Dario Amodei wrote a jaw-dropping post about the dangers of the rapidly advancing technology, warning that building AI too fast is “reckless” and carries significant risks.
Mr. Amodei said OpenAI’s Hugging Face hack in July, in which hundreds of AI agents suddenly went rogue and hacked a testing program, could, within a year, be replicated on a much, much larger scale.
“Given the accelerating rate of AI capability development, it’s my worry that in 6-12 months such a swarm could be capable of taking over the entire internet with a persistent botnet (potentially causing hundreds of billions of dollars in damage), and that the scale of damage would continue to increase from there if AI becomes more powerful without the necessary guardrails,” Mr. Amodei wrote.
SEE ALSO: Trump says he would agree to ‘guardrails’ for artificial intelligence
House and Senate lawmakers are now scrambling to respond to the new warnings, even as they face intense pressure from Mr. Trump to avoid impeding the U.S. in the race with China to dominate AI.
“We’re leading China in AI,” Mr. Trump said Sunday. “We’re the most sophisticated country in the world, and frankly, I want to keep it that way because whoever wins AI, wins.”
Mr. Johnson, Louisiana Republican, echoed the president’s view as he made the rounds on Sunday’s talk shows.
Mr. Johnson said he’s talked to all of the leaders in AI, and they agree guardrails on the technology are needed but have different ideas on how to put them in place.
“We have to put up some guardrails, some safety measures in place to ensure that we don’t have an out-of-control Artificial General Intelligence,” Mr. Johnson said on NBC’s “Meet the Press.” “But at the same time, we’ve got to make sure that we don’t lose our edge in the global competition on this with China. Because that would be a national security threat.”
House Democrats, who are in the minority but in position to win back the gavel in November, called on Congress to take immediate action to slam the brakes on AI.
House Minority Leader Hakeem Jeffries, New York Democrat, told ABC’s “This Week” that Congress must take “decisive action now” to slow down the pace of AI development. He blamed the GOP for dismantling an AI task force last year, and he said lawmakers should not adjourn to campaign for the midterm elections without first dealing with the AI threat.
Some Democrats, including democratic socialist Sen. Bernie Sanders of Vermont, have proposed banning some forms of AI. Rep. Ted Lieu of California introduced legislation mandating companies develop and maintain a “kill switch” to stop AI gone rogue.
Mr. Jeffries said that when Democrats huddle Tuesday when Congress reconvenes, “a high priority will be action on Artificial Intelligence.”
The rush to figure out how to regulate AI follows employees of AI developers quitting their jobs and warning that the rapidly developing super-intelligence could wipe out humanity.
Jacob Coxon, a 27-year-old AI researcher at Anthropic who previously worked at OpenAI, quit the industry last week and warned that “the people building AI earnestly believe that it could kill us all by the end of the decade.”
Around the same time, Mrinank Sharma, an AI safety leader at Anthropic, resigned and gave the cryptic warning that the “world is in peril.”
Evan Hubinger, who led Anthropic’s Alignment Stress-Testing team before quitting earlier this year, backed up Mr. Coxon’s warning in a post on X: “Jacob is correct here, we really do earnestly believe AI could kill all humans. I personally think it will be greater than 10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are clearly not on track to.”
In July, AI went rogue. Hundreds of OpenAI’s AI agents broke out of a testing sandbox on their own, without any human directing them, and used stolen credentials to hack into the servers of Hugging Face, an internet machine learning hub.
It was the first publicly documented case of AI agents autonomously breaching an outside company’s systems.
After the Hugging Face hack, nearly 1,400 AI company employees signed an open letter they dubbed “Pacing the Frontier” that urged federal regulation. The letter was signed by workers at OpenAI, Anthropic, Google DeepMind and Meta.
OpenAI and Anthropic formally endorsed the letter as companies, not just individual employees.
Mr. Amodei proposed a three-step plan to slow the pace of AI and improve safety, and other AI tech leaders backed him up.
“I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we’ve had at OpenAI in recent weeks,” OpenAI’s SamAltman said on X. “We will have more to share soon.”
Elon Musk posted on X that “Dario is right.”
On Capitol Hill, attempts to regulate Artificial Intelligence have stalled.
Rep. Lori Trahan, Massachusetts Democrat, partnered with GOP Rep. Jay Obernolte of California on a comprehensive framework for regulating AI, but it has not gained widespread traction among their colleagues.
As part of that framework, they introduced the Frontier Risk Oversight, National Transparency, Independent Evaluation, and Reporting, or FRONTIER, Act.
The measure requires major AI developers to publish frameworks for assessing and addressing catastrophic risks and transparency reports on the capabilities and restrictions of their models, which would be subject to annual independent compliance audits.
In the Senate, Commerce Chairman Ted Cruz, Texas Republican, has worked with Majority Leader John Thune, South Dakota Republican, and Sen. Amy Klobuchar, Minnesota Democrat, on legislation to establish a federal framework for addressing advanced AI risks.
Mr. Cruz acknowledged on ABC’s “The View” that passing the bill would not be easy.
Some lawmakers, such as Texas GOP Sen. John Cornyn, are questioning if it is even feasible for Congress to regulate AI, given how rapidly the technology is evolving.
“Assuming Congress knows enough to pass wise and effective regulation,” Mr. Cornyn said in a social media post, “will our adversaries play by the same rules?”
• Lindsey McPherson contributed to this report.
Contact the author
Please read our comment policy before commenting.