$10M to End the Career of One AI Regulator

Introduction

An AI super PAC put $10 million on the table to end one man's political career. His crime: he wrote the law that makes AI companies tell you when their models could help build a weapon. The ads attacking him never mention AI. They're about ICE and Sam Bankman-Fried.

The man is Alex Bores, a state assemblymember in the Democratic primary for New York's 12th Congressional District, which votes June 23. Back in November 2025, an AI industry super PAC named him its "number one target," three weeks after he announced. By this week it had spent $7.6 million against him, per the AP. Winning this one seat was never the real goal. The money is there so every other state legislator in the country can see what regulating AI costs you.

The bill that started it

The thing Bores did to earn that campaign is sitting in the public record as enacted law. He co-wrote New York's RAISE Act with State Senator Andrew Gounardes, and Governor Hochul signed it December 19, 2025. It makes large frontier AI developers publish a safety framework and report a critical safety incident within 72 hours, and it defines "catastrophic risk" as a foreseeable, material risk of death or serious injury to dozens to 100-plus people (the threshold shifted during negotiations), or more than $1 billion in property damage. The scenarios it names are the ones that sound like science fiction until you read them in a statute: an AI giving expert-level help with a chemical or biological weapon, an AI running a cyberattack with no human at the wheel, or one slipping the control of the company that built it.

That's the final law, and it got a lot weaker on the way there. The original bill required independent third-party audits, a provision the industry got softened in negotiations. The penalties started at $10 million for a first violation and $30 million after that, and came out the other side at $1 million and $3 million. Both OpenAI and Anthropic said they supported the watered-down version Hochul signed.

So the industry won the negotiation, then went after the man who wrote it. Leading the Future is the super PAC funded by OpenAI president Greg Brockman, Andreessen Horowitz founders Marc Andreessen and Ben Horowitz, and Palantir co-founder Joe Lonsdale, and it announced it would end Bores' career before he could carry the same idea to Congress. Lonsdale put the rationale on CNBC in November: "A random legislator in New York state is not going to stop the AI apocalypse. They're only going to harass and slow us down, and make us lose to China."

Following the $7.6 million

Follow the money and the structure tells you who built it. Leading the Future raised $75.5 million through March 31, almost all from individuals. Brockman and his wife Anna put in $25 million (they gave a separate $25 million to Trump's MAGA Inc.); Andreessen and Horowitz put in $50 million between them. Leading the Future then seeded a Democratic-primary arm called Think Big PAC with exactly $10 million transferred from the parent. Think Big's FEC filing through April 30 shows it raised $10,508,008.58 and had already spent $7,205,188.17, with the AP putting the total against Bores at $7.6 million as of this week.

Now look at the ads that money bought, because none of them say "AI." Think Big's first major ad hit in December: "ICE is powered by Bores' tech. Manhattan is smarter than that." It's a reference to Bores' old job at ICE contractor Palantir. Read that again with the funding in mind: the ad attacks Bores for having worked at Palantir, and it's bankrolled by a PAC whose donors include Palantir's own co-founder. Later ads tied Bores to Sam Bankman-Fried's 2022 political donations. Another warned voters to "know the truth about who is really backing Bores," a swipe at his Anthropic-aligned support that still never names the actual fight.

There's a quieter operation underneath the TV money. Wired reported in May that a dark-money nonprofit linked to Leading the Future paid TikTok and Instagram influencers $5,000 a video to push pro-AI messaging framing Chinese AI as a national security threat. Internal documents Wired obtained said the goal was to "subtly shift public debate toward deregulation" while "intentionally avoiding technical discussions regarding AI quality or safety." The same instinct runs through the whole operation: spend heavily on AI deregulation, then keep the words "AI deregulation" out of everything the public actually sees.

Why the "billionaires bought the seat" story is wrong

Most coverage stops at the eight-figure number, but the obvious read doesn't survive the math. The detail that snagged me: if AI money were buying this seat, Bores would be losing, and he's running near the front. A rival faction backed by Anthropic-aligned groups and Ripple co-founder Chris Larsen spent more in his defense than the attacks did. Gothamist put it at $9.7 million pro-Bores versus $7.7 million against as of June 16. Total super PAC spending across the race is closing on $30 million, one of the most expensive House primaries of the cycle.

And it may have backfired. When Emerson polled in May, Bores sat at 20, in a statistical tie for the lead with rival Micah Lasher at 22, with 32 percent of voters still undecided. Gothamist put it plainly: by throwing the first punch, the people opposing Bores "may have lifted a once-obscure candidate and made him look like a champion of AI regulation." The $10 million punch turned an obscure assemblymember into the guy the AI industry is afraid of, and he rode it to national fundraising and glowing profiles in the Times and The New Yorker. Lasher leads the prediction markets, with Polymarket putting him at 64 percent to Bores' 35 as of June 17, but a betting favorite in a four-way race this undecided is not a sure thing.

So if the money can't win the seat, and might be helping your target, why spend it? The payoff is the price tag itself, posted in public where every other state legislator can read it. Larsen said his own counter-spending "resulted directly from OpenAI's threats to make examples of candidates who seek common-sense regulation." Bores says they're "targeting me to make an example of me." The New Yorker described the spending as meant to set an example for any candidate with regulatory ambitions.

Who benefits

Two eight-figure factions are using one primary as a proxy, and they want different things from it.

The Leading the Future side, the a16z and OpenAI-aligned donors, gets two payoffs if Bores loses. The first is the deterrence signal: a state legislator who wrote a serious AI law gets his career ended by the industry's own money, and the next legislator thinking about a bill watches it happen. The second is federal, and it's where the deterrence signal turns into real money. The House is drafting a three-year preemption of state AI development laws that would wipe out RAISE Act compliance entirely. That compliance is the cost the labs are trying to shed: the 72-hour incident reporting window, the requirement to publish a safety framework at all, the documentation that turns an internal failure into a public disclosure. A three-year freeze makes all of it go away, and a Bores loss engineered by industry money makes the freeze an easier sell by showing state-level AI regulators are politically expendable. The template is Fairshake, the crypto PAC that spent $85 million in the 2024 primaries and reshaped who sits in Congress on crypto regulation. That worked, so the same playbook is now pointed at AI, and Bores is where the industry finds out whether it transfers.

The Anthropic-aligned side wants something subtler: a regulatory moat. Anthropic has built compliance infrastructure and publicly backs state AI rules, so a world where every frontier lab has to publish safety frameworks rewards the lab that already does. Its head of external affairs said the RAISE Act and California's SB 53 "should inspire Congress to build on them." The money traces the same way the other side's does. Anthropic gave $20 million to Public First Action; the Public First network spent more than $6 million backing Bores through a subsidiary. An Anthropic researcher put $790,000 into a pro-Bores PAC, and Larsen, an Anthropic investor, put in $3.5 million. Neither side's corporate name shows up on a single ad, which is exactly how a regulatory war gets fought now.

The part that isn't tidy

This isn't a good-guys-versus-bad-guys story, and the cleanest tell is the donor lists. An OpenAI employee gave $200,000 to a pro-Bores PAC, the opposite of what a clean OpenAI-versus-Anthropic binary would predict. The attack ads sell a tidy story where one company's money is buying the seat and the other side is the resistance, but a donor from inside OpenAI writing a six-figure check for the man OpenAI's president is spending millions to defeat doesn't fit that story at all. Bores' own funding isn't local either: only 12 percent of his money came from inside NY-12. His rival Lasher also supported the RAISE Act, and Brad Carson of Public First told the AP, "He's very good on AI issues too. We win either way." OpenAI formally denies any corporate involvement, which is technically true: Brockman's millions are personal, not corporate treasury money, though he described them publicly as "in service of OpenAI's mission."

For all the proxy-war money, the people voting are the ones left in the dark. A voter in NY-12 who watched every ad in this race would have learned that Bores once worked at Palantir and that a crypto criminal liked him in 2022. They would not have learned that the most AI-soaked primary in the country is a referendum on whether an AI company has to tell the public when its model could help build a bioweapon. That gap, between what the money is for and what the voter is allowed to see, is the disclosure failure the RAISE Act was written to fix, playing out in the campaign to kill it.

The bottom line

Bores might win on June 23. His side outspent the attacks, and the attacks may have made him a star. But the industry extracted most of what it came for the day it wrote a $10 million check with his name on it, because the lesson lands on every other statehouse whether or not Bores keeps his seat. Pass a serious AI law and the people whose models you regulated will spend eight figures to end you, and they'll do it with ads about something else entirely.

So watch the spread on June 23, then watch the federal preemption talks the week after. If a three-year freeze on state AI laws picks up speed right after the most-watched AI regulator in the country gets buried under $7.6 million in ads about ICE, you'll know what the spending was really buying. The question for the next legislator who drafts a disclosure bill is simpler, and they already have the answer in front of them: how much is your career worth, and who gets to decide?