AI executives and companies are betting that spending millions in the 2026 midterm elections will allow them to influence AI bills being developed in Congress.
As of the end of June, the two biggest artificial intelligence political action committees have dropped at least $44 million into 40 House and Senate candidates, per a CNBC analysis of Federal Election Commission data. That’s an early taste of how the groups will spend the more than $200 million they’ve raised on the rest of primary season and into the general election, according to fundraising totals provided by groups.
The spending by the burgeoning AI industry makes it an increasingly powerful player in the Washington influence space. The companies — through their PACs — are setting themselves up to shape how the first national legislation to regulate AI use takes form.
Brad Carson, who heads Public First Action, a nonprofit organization with several PACs, said he’s seen more bills introduced and discussion around AI legislation, especially as concerns about the capabilities and risks of powerful AI models like Mythos and Claude Fable have come into the spotlight. While any legislation is unlikely to cross the finish line this year, given the limited number of days lawmakers are in session, both parties have signaled AI will continue to be a priority in coming years.
“They have a lot of benefits. They have a lot of dangers. And you can’t just release them into the wild with no government concern,” Carson told CNBC. “Everybody from the right to the left, from pro-Trump to anti-Trump recognizes that.”
Josh Vlasto, co-leader of Leading the Future, said coming up with the right regulatory structure is critical for lawmakers.
“It is so important that we do this now and urgently, because it is still the early innings of the technology, but it is being adopted quickly, at scale,” he told CNBC.
The overwhelming number of candidates supported by both PACs have won their primaries. Of the 28 candidates Leading the Future has backed, 25 have won their primaries, two have yet to face their elections and only one — Jesse Jackson Jr. — lost. The group also opposed Alex Bores, who lost the Democratic primary in New York’s 12th Congressional District.
Public First Action has backed candidates in 11 races. With the exception of Bores, every candidate it has supported has won. Carson said the group plans to spend in 50-60 races by the end of the midterms.
The playbook AI companies are using isn’t a new one. In the 2024 elections, crypto-backed PAC Fairshake dropped an eye-popping $200 million into elections, supporting pro-crypto candidates on either side of the aisle. The result: A major bill on stablecoins became law, and significant progress was made on a rules-of-the-road digital assets bill favored by major crypto companies like Coinbase and Ripple.
So far, Leading the Future has spent more than $24 million on primary races through the end of June, according to data filed with the Federal Election Commission. The group said it raised $125 million by the end of 2025, in part from donors including private equity firm Andreessen Horowitz, Open AI co-founder Greg Brockman, Palantir co-founder Joe Lonsdale, SV Angel founder Ron Conway and AI software company Perplexity.
Public First Action, which launched last year, has spent $20 million so far, and it announced last month it had raised $80 million through the end of June. The group received $20 million from Anthropic, although the amount is restricted to educating the public on AI policy and not for political purposes, per a spokesman for the PAC.
Pubic First Action doesn’t disclose its donors; Anthropic disclosed its own donation. But Carson said the group has received donations from employees of OpenAI, Google, DeepMind and X.
Two AI groups spend big in politics
The two groups have butted heads several times in races, spending against each other in the Manhattan Democratic primary and even trading swipes in interviews.
Yet the policy differences between the two are far more nuanced than “pro” or “anti” regulation. Both groups back some degree of guardrails and even overlap in areas like the need to protect children online.
The biggest differences touch on one of the stickiest issues in Washington — whether a single federal standard should preempt state laws around AI. But even on that issue, the groups aren’t fully on opposite ends of the debate.
Leading the Future advocates for a “broad, national, consistent framework for regulation governing AI,” Vlasto said in an interview with CNBC. He denied the group was against state laws, pointing to the its support of New York’s landmark AI law, the RAISE Act, which Bores helped lead as a New York assemblyman.
But the RAISE Act shows how complex the group’s stance is. Leading the Future spent about $8 million opposing Bores in large part because of his push for a more aggressive RAISE Act than was eventually signed into law.
Before the bill was signed into law, New York Gov. Kathy Hochul successfully pressed lawmakers to agree to changes to weaken reporting requirements for AI companies and the size of penalties — bringing the New York law more in line with one in California. Those changes resulted in Leading the Future supporting the final law while still opposing one of the lawmakers who backed the earlier version of the bill.
Public First Action is more supportive of state laws and has fought efforts to preempt them, although Carson said if Washington can come up with a “comprehensive federal approach to these problems, then preemption is a natural part of our constitutional order.”
Republicans on Capitol Hill have tried, and failed, several times to preempt state laws. House Majority Leader Steve Scalise, R-La., told CNBC that state laws are “hurting innovation” and overriding them is “going to be the foundation of anything we do.”
Rep. Ted Lieu, D-Calif., a co-chair of a commission on AI established by House Democrats, said that while there is “definitely bipartisan disapproval of preempting with nothing,” he noted many Democrats recently backed a kids online safety bill that set a federal standard for privacy standards as a floor.
