Congress has no shortage of ideas for regulating the artificial intelligence industry but lacks consensus on how to protect humanity from the risks of AI systems operating beyond human control.
Lawmakers say they want to ensure that whatever they do does not stifle technological progress or cause China to surge ahead of the U.S. in the AI race.
President Trump is not interested in new federal regulations for that reason and has said the recent panic over AI operating beyond human control and wiping out humanity is a “hoax.”
Yet he has agreed to hold a meeting with AI leaders “within the next week,” said House Speaker Mike Johnson, Louisiana Republican, who discussed the matter with the president Monday night.
Mr. Johnson said the meeting is intended to discuss a “responsible, balanced approach” to preventing AI from escaping human controls through a combination of developer self-policing and federal regulations.
“What the president is saying is you’re not all going to be dead in 10 years,” Mr. Johnson said. “That’s a hoax.”
The speaker said one reason for gathering AI leaders is that Congress does not have the latest information on frontier models still in development or on how swarms of AI agents have executed cyberattacks outside their human instructions.
In the meantime, lawmakers in both chambers are resurfacing and revamping old AI regulation bills and introducing new ones to address growing concerns from industry leaders and the public about AI systems operating beyond human control.
AI kill switch
Sen. John Kennedy, Louisiana Republican, said he plans to go to the Senate floor Wednesday and seek unanimous consent to advance a bill that would require every AI developer to build kill switches for their models.
He said the measure does not create government regulations, which the president opposes, but addresses the risk that advanced AI models could become “the terminator.”
Reps. Ted W. Lieu, California Democrat, and Nathaniel Moran, Texas Republican, have a bill called the AI Kill Switch Act that would not only require AI developers to maintain a shutdown mechanism but also authorize the Department of Homeland Security to order the shutdown of an AI system that could cause catastrophic harm.
’Light touch’
Senate Commerce Committee Chairman Ted Cruz, Texas Republican, said he wants to hold a markup this month on bills to regulate AI, but that will depend on getting bipartisan agreement.
In particular, he wants to consider a revamped version of a bill the committee advanced last Congress, led by Sens. John Thune, South Dakota Republican, and Amy Klobuchar, Minnesota Democrat.
Mr. Thune said Tuesday that he views the bill as a “light-touch approach” to federal regulations that has previously had bipartisan support but is being blocked by Democrats.
As written in the 119th Congress, the Artificial Intelligence Research, Innovation, and Accountability Act directed the National Institute of Standards and Technology to develop “methods to detect and mitigate anomalous behavior” of AI systems.
It also required AI companies to submit risk assessments to the Commerce Department before releasing new models and to provide regular transparency reports after deployment.
One of the hang-ups now is that Republicans are looking to add language that would preempt states from enacting or enforcing similar regulations, which is a nonstarter for many Democrats.
Independent monitors
An idea with broad support among both parties is to require independent auditors to monitor the safety of AI models.
Mr. Johnson endorsed that approach, and Sen. Adam B. Schiff, California Democrat, said requiring truly independent monitors is “one of the most promising suggestions” for regulating AI.
“Legislation is necessary to make this a requirement of all companies, not simply those who agree to do so — to ensure the [independence] of these monitors and that companies properly and safely pace new developments,” Mr. Schiff said.
That is one of the features of the Frontier Risk Oversight, National Transparency, Independent Evaluation, and Reporting, or FRONTIER, Act.
The bill from Reps. Jay Obernolte, California Republican, and Lori Trahan, Massachusetts Democrat, requires every large AI developer to retain a government-licensed independent verification organization, or IVO, to conduct risk monitoring.
OpenAI has endorsed that piece of the FRONTIER Act, but not the full bill.
Federal monitors
Some lawmakers do not believe independent monitoring is enough.
“With our cybersecurity, biosecurity, critical infrastructure, and national security on the line, every AI developer should have to submit their frontier models to the U.S. government for a shot-clocked review before they release them to the public,” said Rep. Josh Gottheimer, New Jersey Democrat, citing the need for the security clearances and compute power needed for effective testing.
Sens. Josh Hawley, Missouri Republican, and Richard Blumenthal, Connecticut Democrat, pointed to their bill, which would require the Energy Department to establish an evaluation program to review frontier AI models.
Mr. Blumenthal said the goal is a government review process that can assess risks as the Food and Drug Administration does for pharmaceutical drugs.
“We worked on this bill with a number of AI scientists and others within the government who have done this in the past, and what they recommended was the DOE has sort of the infrastructure in place,” Mr. Hawley said. “It doesn’t have to be there. … We just want to make sure that it’s a department that has capability to assess these frontier models for national security.”
Sen. Mark R. Warner, Virginia Democrat, has legislation to establish an AI risk board within the National Institute of Standards and Technology to conduct such assessments.
Moratoriums
Some Democrats are pushing for temporary or permanent moratoriums on the development of advanced AI models.
Sen. Bernard Sanders, Vermont independent, said he is planning to introduce legislation with Rep. Greg Casar, Texas Democrat, to permanently ban the development of artificial superintelligence, which he described as “an AI mind smarter than any human being and capable of operating independently beyond our control.”
“Like nuclear weapons, artificial superintelligence is a threat to humanity — and it must be treated as such,” he said. “Further, this legislation would put an immediate pause on advanced AI development until we have clear safety rules in place, developed by the most knowledgeable scientists in our country.”
Mr. Johnson dismissed the calls for an AI moratorium as a “stupid, knee-jerk reaction” that would cause the U.S. to lose its edge over China and lead to “serious national security implications for every American family.”
Data centers
Separate from the AI safety debate, Congress has worked on legislation to ensure local communities do not have to shoulder the energy and infrastructure costs of AI data centers.
The House was set to vote Tuesday night on the bipartisan Ratepayer Protection Act, which encourages state regulators to adopt rules requiring data centers with an electricity demand of 100 megawatts or more to pay for the generation, transmission and other infrastructure necessary to meet that demand, so that the costs are not passed on through higher utility rates.
“As the United States competes to win the global AI race and outcompete adversaries like Communist China, we must ensure that growth does not come at the expense of hardworking families,” said Rep. Gabe Evans, Colorado Republican and the bill’s lead sponsor.
Mr. Thune said Senate action on the bill before the election would depend on whether they could get unanimous consent from all 100 senators to hold a quick vote, as there is not enough time to run through all the normal Senate procedural hurdles.
Contact the author
Please read our comment policy before commenting.