Manhattan's Democratic primary for Rep. Jerry Nadler's seat drew more than $40 million in outside spending, with a pro AI super PAC network spending about $8 million to defeat AI safety candidate Alex Bores.
Silicon Valley's first major political bet on a single congressional candidate is over, and the lobby got what it paid for. Roughly $8 million from a pro-AI super PAC network spent the past few months trying to bury New York Assemblymember Alex Bores in attack ads, and on Tuesday Bores lost the Democratic primary for Rep. Jerry Nadler's Manhattan seat to fellow assemblymember Micah Lasher. Bores had made two specific AI policy proposals the spine of his campaign: a state-level RAISE Act requiring frontier AI developers to publish and follow formal safety protocols, and a federal "AI Dividend" sending cash payments to Americans harmed by AI-driven job displacement.
That $8 million is one slice of a much larger outside spending war. Outside groups poured more than $40 million into NY-12 across the eight-candidate Democratic field, making this the first US congressional primary organized almost entirely around AI policy. On the opposing side, an AI-safety network anchored by an Anthropic-backed organization, the Dream NYC PAC linked to Anthropic researcher Daniel Ziegler, and the You Can Push Back fund linked to crypto billionaire Chris Larsen spent roughly $20 million defending Bores. Gizmodo first framed the race as the AI Super PAC's first major target; CNBC reported the dollar totals this week.
The money traces a specific architecture. Leading the Future, the pro-AI super PAC network behind the spend, was formed in August 2025 and is backed by Andreessen Horowitz, OpenAI president Greg Brockman, Palantir co-founder Joe Lonsdale, and Perplexity, according to Wikipedia's Leading the Future entry, which cites FEC filings, the Wall Street Journal, and the Washington Post. The network has raised more than $140 million since launch, per FEC data mirrored by Transformer's PAC tracker, and operates a dark-money nonprofit affiliate called Build American AI run by Nathan Leamer. The Think Big PAC, the specific entity that ran anti-Bores advertising, is a Leading the Future affiliate. A counter-PAC called Public First was founded by former members of Congress Brad Carson and Chris Stewart explicitly to oppose the network.
Bores was not a fringe candidate. He had carried the RAISE Act through the New York State Assembly, the first state-level bill requiring leading AI developers to publish and follow formal safety protocols, and was running to extend a federal version alongside the AI Dividend in Washington. Lasher, his winning opponent and a RAISE Act co-sponsor, is heavily favored to hold the safely Democratic seat in November. According to Bores's Wikipedia profile, his policy platform was unusually specific for a primary: not "AI regulation" in the abstract, but named protocols, named reporting obligations, and a named redistributive mechanism.
Bores framed his concession on X as evidence that demand for AI policy accountability is real, not diminished. "They set out to make people afraid to stand up to them," he wrote. "Instead, they learned just how ready people are to push back." Lasher's victory clip, a YouTube Shorts clip flagged by Gizmodo, struck a different note, rebuking unnamed "big AI companies" for their unusual interest in a state-level primary. Lasher, not Bores, is now the candidate carrying the RAISE Act into the general election. If he holds that position, the pro-AI super PAC network that opposed Bores will spend the next five months fighting either for or against Lasher on the same policy architecture.
What happened in NY-12 is closer to a structural test than a verdict. The pro-AI super PAC network spent its first congressional primary at scale, against a single named target, with about $8 million of attack advertising routed through a Leading the Future affiliate, and won on points. The AI-safety coalition spent roughly $20 million defending one candidate and lost, but did so while moving a specific governance model, binding safety protocols plus redistributive cash transfers for automation displacement, from advocacy into an electoral position with named sponsors on both sides. That position now has a defeated-but-active campaign operation behind it, a safely Democratic incumbent in November who has co-sponsored one of its core bills, and roughly $70 million of additional industry money waiting in the network's broader war chest, according to NOTUS.
A few caveats are worth holding. "First major target" is Gizmodo's framing, not an established structural claim. The Campaign Legal Center filed an FEC complaint in May alleging shell-company transparency evasion by Leading the Future affiliates; the complaint is unresolved. OpenAI has publicly distanced itself from Brockman's personal donations, according to Business Insider, so the network's institutional posture and its president's personal spending should not be conflated. Exact vote margins and final district-wide returns were not in the publicly available reporting at the time of writing and should be checked against the AP or New York Times results wire before any second-day coverage runs.
What to watch next: whether Lasher keeps the RAISE Act co-sponsorship through the general election, whether the AI Dividend gets reintroduced as a federal bill, and whether the pro-AI super PAC network names a different primary target for the next cycle. The donor architecture exists on both sides. What is unresolved is whether Lasher carries the safety architecture into November or routes around it.