Reps. Brett Guthrie, R-Ky., and Ted Lieu, D-Calif., said on Oct. 8 that slowing the development of artificial intelligence (AI) is not a viable option, but Congress needs to establish federal safeguards to address the technology’s potential risks.
Speaking at the Punchbowl News Tech Summit, both lawmakers emphasized the need to balance AI safety with U.S. competitiveness, pointing to bipartisan legislation that would allow dangerous AI systems to be shut down without restricting AI development.
Guthrie said a global pause in AI development would be difficult to enforce and could leave the United States behind China.
“China’s not going to stop, and so we have to say, all right, what’s the immediate threat?” Guthrie said.
Guthrie said he opposes both congressional inaction and a moratorium on AI development. He urged lawmakers to focus on specific risks, including cybersecurity and bioterrorism, while establishing guardrails for the technology.
Lieu agreed that slowing AI development is not an option but noted that the lack of a comprehensive federal AI law has left states to develop their own regulations.
Several states have moved to establish AI safeguards in recent weeks. Maryland Gov. Wes Moore announced a framework in September addressing advanced AI systems, worker protections, and children’s safety. On Sept. 23, Oregon Gov. Tina Kotek ordered the development of AI procurement safeguards, including independent safety reviews and consideration of a “kill switch” for certain systems.
While Lieu said he does not oppose state-level AI regulation, he emphasized the need for a consistent federal framework.
Lawmakers back AI “kill switch”
Both lawmakers also voiced support for requiring AI developers to maintain a “kill switch” that would allow dangerous AI systems to be stopped if they pose a threat.
“If it goes rogue, you can stop it. [We need] to make sure they have those [kill switch] in place to do it,” Guthrie said.
Lieu pointed to the AI Kill Switch Act, which he introduced this summer with Rep. Nathaniel Moran, R-Texas, as an example of a targeted safety measure that would not dictate how AI models are built or prevent their development.
Under the legislation, covered AI developers would be required to maintain the technical ability to slow, suspend, or fully shut down certain AI systems.
The bill would establish a graduated response framework that allows the government’s response to match the severity of an incident, ranging from slowing a system to shutting it down entirely.
The legislation would also require incident reporting and the preservation of forensic records to help officials investigate failures.
“If it goes rogue or if it’s got capabilities you did not expect, you got to be able to turn it off,” Lieu said.
Congress continues to weigh AI regulation without agreement on what a comprehensive federal framework should include.
Guthrie said lawmakers could pursue bipartisan AI legislation during a lame-duck session following the November elections, before newly elected members of Congress take office on Jan. 3.
“Hopefully, in a lame duck, we can get something done,” Guthrie said, emphasizing that any legislation would need bipartisan support.
Lieu said he also sees an opportunity for bipartisan action on targeted AI legislation during a lame-duck session but expressed skepticism about Congress passing a comprehensive AI package.