AI has become one of the better-funded causes of the 2026 US midterms. A CNBC analysis of Federal Election Commission filings through October 20, 2026, summarized by AI Weekly, puts AI-focused political groups at about $78.5 million on federal races, with nearly three-quarters spent before the general election began. A Techmeme-linked piece from Puck's Ian Krietzberg frames it as nearly $80 million, with pro-regulatory groups outspending the network around Leading the Future.
We could not open the Puck piece, so this post rests on the CNBC analysis and earlier reporting, and it flags where the numbers differ.
TL;DR: the numbers
| Item | Figure | Source |
|---|---|---|
| Total AI-focused political spending, federal races | About $78.5M through Oct 20 | CNBC analysis of FEC filings |
| Share spent in primaries | Nearly three-quarters | Same |
| Leading the Future-aligned groups | About $26M primaries plus about $8M general (about $34M) | Same |
| Safeguards-focused groups | About $32.1M primaries plus $6.9M general (about $39M) | Same |
| Anthropic to Public First Action | $20M in February, $20M in July | Same |
| Earlier tally | More than $50M | Transformer, cited by Fortune in June |
The two camps
Leading the Future and allies favor faster AI development and lighter, more uniform federal rules, which in practice means limiting the patchwork of state laws. Per the CNBC summary, its named donors include Andreessen Horowitz, OpenAI co-founder Greg Brockman, Joe Lonsdale, Ron Conway and Perplexity.
Safeguards-focused groups, including Public First Action, want stronger protections and more room for states to regulate. Public First Action head Brad Carson said Leading the Future "is driven by three billionaires who are close to Donald Trump" and that AI regulation should be "more democratically accountable." That is one side's characterization of the other, not a verified fact.
The groups did not just spend in parallel. Leading the Future, Public First Action and Innovation Council Action backed opposing candidates in at least six primary races, per the summary. One example: Think Big, a Leading the Future affiliate, spent $310,000 supporting Democrat Jared Moskowitz in Florida's 25th District primary.
Who outspent whom?
On the CNBC figures the safeguards side is ahead, about $39 million to about $34 million. That is a modest lead, and it depends on which groups are assigned to which camp. Several groups spend on ads that are not about AI at all; an NBC News piece headed "Why ads funded by AI groups aren't about AI" makes that point, so the money is a measure of political investment, not of how much voters heard about AI.
The Puck framing, as shown in the headline, says pro-regulatory groups outspent Leading the Future's network. That agrees in direction with the CNBC figures but we have not verified Puck's numbers.
Anthropic's money and the "restricted" claim
Anthropic committed $20 million to Public First Action in February and another $20 million in July. Anthropic says the funds are restricted to policy and public education and cannot be used for election ads. That is a company statement. Public filings can show where Public First Action's own spending went, but whether restricted money and electioneering money stay separate is an accounting question outsiders cannot settle from headlines.
For background on Anthropic's wider political activity, see our reporting on its presidential engagement program for 2028 candidates, and for OpenAI's side, OpenAI's election war room.
Why this matters to people who build with AI
The money is a fight over who writes AI rules. If a federal framework preempts state laws, builders get one set of obligations; if states keep their room, they face a patchwork. Either outcome changes compliance work, disclosure requirements, liability and what frontier labs can ship. We covered the Democratic side of this in Democrats, AI regulation and the midterms, and how AI infrastructure has become a ballot issue in data center backlash in midterm races and the Maine data center revolt.
A timeline of how the numbers grew
Pieced together from the reports cited above, the cycle looked like this:
- February 2026: Anthropic commits $20 million to Public First Action, per the CNBC-based summary.
- June 2026: Fortune reports on a proxy fight in New York's 12th District, framed as Anthropic and OpenAI-linked money on opposite sides with no clear winner, and cites a Transformer tally of more than $50 million across AI-focused groups.
- July 2026: Anthropic commits a second $20 million.
- October 20, 2026: FEC filings analyzed by CNBC show about $78.5 million, nearly three-quarters of it spent before the general election.
- October 30, 2026: Puck, per a Techmeme headline, reports nearly $80 million, mostly before the primaries, with pro-regulatory groups outspending the Leading the Future network.
The growth from roughly $50 million to roughly $78 million between June and October is partly new spending and partly the difference between counting methods. Without the underlying filings in one place you cannot split those two, which is why we avoid presenting any single total as settled.
What the money says about the labs
Two of the biggest frontier labs and their backers sit on opposite sides of a policy question that affects both of their products. That is unusual: normally an industry lobbies as a bloc. Here the split is between those who want a single light federal standard and those who want stronger guardrails and state flexibility. Readers should note that donors named for Leading the Future are investors and executives, not the companies themselves, while Anthropic's contribution is a company commitment. The distinction matters when you ask what a lab "wants."
It also shows how AI policy is being fought through ordinary campaign tools: independent expenditures in primaries, affiliates such as Think Big, and ads that, per NBC's reporting, are often not about AI at all. If you follow AI regulation as a builder, the signal to watch is not ad spending but the text of bills and the positions candidates take once elected.
What this means for what you build or pay
For most builders the practical effect is indirect but real. A federal preemption law would simplify compliance for products sold across states, while strong state laws mean disclosure, safety testing or liability rules that differ by location. If you ship AI features to US users, keep a short list of state AI laws that touch you, build disclosure and logging in from the start, and avoid designs that only work under the lightest rules. Costs from rules show up later as legal review, audit work and slower launches, not as a line in a campaign filing.
Caveats when you read these totals
- Counting differs. Totals depend on which committees are included and the date range. Fortune cited more than $50 million in June; a New York Times tally cited elsewhere was at least $55.7 million; CNBC's October figure is about $78.5 million. These are not contradictions, they are different snapshots.
- Independent expenditures versus contributions. Spending on ads and mail is reported differently from donations to committees. A group can raise money that is not yet spent.
- Dark money. Some political nonprofits do not disclose donors, so FEC data understates the total.
- Spending is not outcomes. We have no evidence here on whether the money changed any result.
- Filing lag. Reports through October 20 will be updated before and after election day.
Questions readers are asking
Is this the AI industry buying the election? The filings show large spending, but they do not show that it determined any result. Spending in primaries is also split between opposing camps, which partly cancels out.
Why would a lab fund a safeguards group? Anthropic has framed its giving as support for policy and public education. Critics of any lab's political giving argue it can shape rules in the donor's favor, and supporters argue it balances other industry money. Both claims are arguments, not findings.
Does the money buy AI ads? Often not directly. Reporting cited by NBC says many ads funded by AI groups are about other issues, because candidates win on cost of living and local concerns, not on model policy.
Where do I find the raw data? The FEC database lists each committee's receipts and independent expenditures, and committee websites list their stated positions.
How to check the numbers yourself
Search the FEC's public database for each committee by name, filter independent expenditures by date, and compare to what the committee's own site says. Use the same date cutoff when comparing any two outlets. If you cite a total, name the outlet and the cutoff date, as we have done here.
What to watch next
Post-election filings will show final totals and which races the money touched. Watch whether federal preemption language appears in must-pass legislation after the election, whether either camp expands into 2028 planning, and whether the labs behind the donations change their public positions. For readers following our state of AI policy coverage, this is a useful moment to separate lobbying money from legislation that actually moves.
Related reading
- OpenAI's election war room: what is confirmed
- Anthropic's presidential engagement program
- Democrats, AI regulation and the midterms
- Data center backlash in midterm races
- Maine data center revolt and the Senate race
- Alexandr Wang, Musk and AI legislation
Figures are as reported on October 27 to 30, 2026 from FEC-based analyses and will change as new filings arrive.