Getting your
Trinity Audioplayer ready... In 2024, at the dawn of the modern artificial intelligence age, Rep. Jay Obernolte, R-Hesperia, was optimistic that Congress could come together in a bipartisan fashion and create a legislative framework for responsible AI development that would benefit all Americans.
Two years later, companies have laid off workers with the intention of replacing them with AI. Americans of all political persuasions oppose AI data centers in their communities. AI has escaped containment and hacked into other companies’ computers. AI researchers say there’s a 10% chance it will wipe out humanity by the end of the decade. And the leaders of America’s largest AI companies are collectively calling for the industry to slow down its pace of development, in order to create better guardrails on the technology, something President Donald Trump has scoffed at.
But what there hasn’t been is a lot of progress on the legislative front.
“It’s been frustrating how slowly it’s going on,” Obernolte said on Friday, Sept. 18. “The need is urgent.”
In 2024, he chaired the bipartisan House Task Force on Artificial Intelligence. In December of that year, the task force’s final report outlined an optimistic future for AI’s impact.
“AI has tremendous potential to transform society and our economy for the better and address complex national challenges,” the introduction to the report reads in part. “From optimizing manufacturing to developing cures for grave illnesses, AI can greatly boost productivity, enabling us to achieve our objectives more quickly and cost-effectively. Nevertheless, we also recognize that AI can be misused and lead to various types of harm.”
The report offered suggestions on guardrails for its development and use, providing a “targeted approach that balances the need to promote vibrant AI innovation while safeguarding Americans from potential harms as we enter an era of widespread adoption of AI,” the 273-page report reads in part.
The report offered 84 “concrete policy recommendations” agreed upon by a bipartisan group of 24 Representatives.
“That was proof that AI regulation doesn’t have to be partisan,” said Obernolte, who today is the chairperson of the House Subcommittee on Research and Technology.
Making AI regulation bipartisan is Obernolte, who has a strong track record of working across the aisle going back to his days in the state Assembly.
“Whatever we create has to be durable,” he said, and not subject to dramatic changes whenever different parties take control in Washington.
“So far, we have codified one” of the 84 proposals, the Take it Down Act, which created federal penalties for sharing of non-consensual intimate images, including AI-generated deepfakes. The act was signed into law by Trump in May 2025.
“We have quite a few (more proposals) that are queued up for action,” Obernolte said.
In January, he and Rep. Lori Trahan, D-Massachusetts, released a draft of the Great American AI Act, a bipartisan effort to create a federal framework for how the U.S. governs AI.
But it hasn’t been submitted to a bill as-is. Instead, many of the proposals are being split into smaller bills and making their way through Congressional committees on their own.
“AI is so broadly cross-jurisdictional that it does not fit neatly into the committee structure of the house,” Obernolte said.
The GAAIA Act could easily have had to go through eight different committees if it had to be considered as a single act, he said. Instead, Congressional committees are tackling smaller pieces relevant to their areas of expertise, with the goal of recombining everyone back together, like the Power Rangers forming up to create Megazord.
“The goal is to reassemble it into one bill and that will be available for floor action by the end of the year” after Congress returns from the election season recess that began Friday, Obernolte said.
In the meantime, he welcomed the public’s growing interest in artificial intelligence.
But, “I think it’s important to be clear-eyed about what the dangers are,” he said.
The public has been influenced by decades of science fiction “disinformation,” according to Obernolte.
“I am not worried about something out of the Terminator movies where an army rises up to kill humanity in the middle of the night,” he said.
Instead, Obernolte worries about people using artificial intelligence for malicious ends, whether that’s criminals, hostile nations or terrorists.
“When we devise a regulatory framework, it has to be based on mitigating those risks — the real risks,” he said. “If we were worry about the army of evil robots — that aren’t real — we run the risk of overregulating AI.”
He laughed at the idea that there’s a 10% chance that humanity will be destroyed by artificial intelligence, something Trump has scoffed at as well.
“He calls that a hoax,” Obernolte said. “I happen to agree with him is that these existential risks are exaggerated.”
The specific threat former Anthropic researcher Jacob Coxon was warning about on social media was about recursive self-improvement. That’s a technique where AI develops its own improvements, with little to no human oversight. That’s something that Anthropic has just started doing and is not yet a common practice, according to Obernolte.
“The other thing the White House is worried about is the risk of over-regulation,” said Obernolte, who studied engineering and applied science at Caltech before pursuing a master’s degree in artificial intelligence from UCLA. “China would like nothing better than for us to squander our lead. That’s what you’re hearing coming out of the White House and I don’t think that’s incompatible with what we’re trying to do.”
The current conversation leaves out the overwhelming promise of artificial intelligence, he said. “AI is already the greatest tool for the dissemination of human knowledge that mankind has already invented,” Obernolte said. “I don’t think people have thought about the profound implications for human advancement that has enabled. And that’s today. But AI will shortly be the most powerful tool for humankind has ever invented.”
Artificial intelligence has solved math problems that have baffled mathematicians for centuries, can now identify pancreatic cancer early enough to treat the deadly disease, detecting cybercrime, and more.
“That’s why this is so important to me, that’s why we did this, that’s why it’s so important we get this balance right,” Obernolte said.