The Democratic primary in New York’s 12th Congressional District ended the way so many modern political contests now do: with a candidate who had been turned into a symbol, a flood of outside spending that made the race feel less like local politics and more like a referendum on an industry, and a result that was close enough to keep everyone arguing about what it “really” means.
Alex Bores, a New York state Assemblyman whose profile rose sharply after he became the target of a pro-AI super PAC, narrowly lost the nomination to Micah Lasher. The outcome means Bores will not advance to represent the district, and it also marks the end—at least for this cycle—of what campaign watchers have described as a $27 million “AI proxy war” between two of the most influential forces in the frontier AI ecosystem: Anthropic and OpenAI.
But calling it a proxy war can make it sound abstract, like a chess match played by faceless strategists. In practice, the contest was intensely personal and intensely local. It was also unusually revealing about how AI policy is being fought in the open: not only through legislation and regulatory frameworks, but through advertising narratives, voter targeting, and the strategic use of candidates as vehicles for broader ideological and commercial goals.
To understand why this race mattered, you have to start with what Bores had already done before the campaign ever became national news.
Before the primary, Bores was known in Albany circles as a lawmaker with a tech background and a willingness to engage directly with the hardest questions around frontier AI. He coauthored and helped pass the RAISE Act, a high-profile effort to establish guardrails and safety requirements for frontier AI companies. A version of his approach was signed into state law last year, giving the bill a kind of permanence that many AI policy proposals never reach.
That matters because the RAISE Act wasn’t just another piece of legislation. It was a statement about what New York intended to demand from companies building systems that could reshape everything from education to employment to national security. It also created a clear point of friction between lawmakers who wanted enforceable safety obligations and the AI industry players who often argue that regulation should be slower, narrower, or handled through different mechanisms.
In other words, Bores didn’t enter the primary as a blank slate. He entered as someone who had already taken a public position—one that could be interpreted as either responsible governance or an obstacle, depending on where you stood in the AI debate.
Then came the outside spending.
According to reporting and campaign finance tracking cited in coverage of the race, the contest drew extraordinary attention for the scale of outside spending. The figure that has circulated widely—$27 million—has become shorthand for the idea that the election was not merely about Bores versus Lasher, but about competing visions of how AI should be governed and who should influence that governance.
The super PAC at the center of the controversy, Leading the Future, has been described as a $100 million operation. Its involvement in the Bores campaign didn’t just add money; it added a narrative. And in modern elections, narratives are often more decisive than policy details, especially when voters are overwhelmed by the sheer volume of messaging.
Bores’ opponents framed the RAISE Act and related guardrails as something that would slow innovation or impose burdens that would ultimately harm consumers and workers. Supporters of Bores, meanwhile, argued that the legislation was exactly what frontier AI requires: concrete safety expectations rather than vague promises.
The result was a campaign that felt like it was happening on two levels at once. On the ground, voters were deciding who should represent them in Congress. But in the background, the race became a stage for a larger argument about whether AI safety should be treated as a public-interest obligation—or as a negotiable constraint that industry can shape through lobbying, messaging, and political pressure.
What makes the Bores story particularly striking is that the campaign did not simply “attack” him. It elevated him.
Targeting a candidate can sometimes backfire, and in this case it appears to have done exactly that—at least initially. Bores’ popularity surged after he became the focus of the pro-AI super PAC. That surge is part of why the primary ended up so close. The more the outside spending tried to define him, the more voters seemed to respond to the underlying question: if someone is willing to push for guardrails, why are powerful interests trying so hard to stop him?
This is one of the paradoxes of proxy politics. When outside groups pour resources into a race, they often assume they are controlling the terms of the debate. But they can also create the conditions for a backlash—especially when the targeted candidate has a record that voters can evaluate beyond the ads.
In Bores’ case, the record was tangible. The RAISE Act wasn’t a slogan; it was a legislative achievement. It gave voters something concrete to weigh: a law designed to require safety and guardrails from frontier AI companies. Even people who disagreed with the approach could see that Bores wasn’t improvising. He had built a policy framework and pushed it through.
So why did he still lose?
The answer likely lies in the difference between being a symbol and being the nominee.
A candidate can become a rallying point, but winning a primary requires coalition-building across a range of voters with different priorities. In a district like New York’s 12th, where Democratic voters often include a mix of progressive activists, pragmatic centrists, and community-focused constituencies, the question isn’t only “Who is right about AI?” It’s also “Who can win, who can govern, and who fits the district’s political temperament?”
Micah Lasher, the candidate who ultimately won the nomination, benefited from the fact that the race’s national framing didn’t fully translate into a durable majority. Outside spending can intensify attention, but it can also polarize. If voters feel that a campaign has been hijacked by distant interests, they may react against both sides of the messaging—even if they agree with parts of the policy debate.
There’s also the structural reality of primaries: turnout patterns can be unforgiving. When a race becomes a national spectacle, it can draw attention from people who don’t normally vote in that primary, while simultaneously failing to mobilize the most reliable local base. The result can be a narrow margin that reflects a complex mix of enthusiasm, skepticism, and fatigue.
In Bores’ case, the surge in popularity after being targeted suggests that the outside attacks didn’t simply crush him. Instead, they may have helped him consolidate a certain kind of support—support that recognized the importance of AI safety and resented the idea that powerful AI interests could dictate political outcomes.
But consolidation is not the same as winning.
The primary ended with Bores narrowly losing, which means the electorate was divided enough that even a strong response to the super PAC’s involvement wasn’t sufficient to carry him over the line. The seat is expected to go to Lasher, who will replace Rep. Jerry Nadler.
That transition is important for another reason: it underscores how quickly AI policy debates can become electoral leverage. Nadler’s tenure and the district’s political history are not reducible to AI. Yet the campaign’s defining feature was AI governance—specifically, the RAISE Act and the guardrails it represented.
This is where the “proxy war” framing becomes more than a catchy phrase. It points to a new pattern in American politics: when a technology becomes politically salient, the fight over its rules can migrate into elections, turning legislative proposals into campaign ammunition.
And once that happens, the incentives change.
For lawmakers, supporting AI safety measures can become a risk if powerful industry actors decide to punish that stance electorally. For candidates, taking a position can become less about policy conviction and more about anticipating how outside groups will interpret and amplify their choices. For voters, the debate can become harder to evaluate because it arrives packaged in advertising narratives rather than in hearings, technical briefings, or transparent regulatory processes.
In theory, elections are where democratic accountability happens. In practice, when outside spending dominates, accountability can become diffuse. Voters may feel they are choosing between two candidates, but the real contest may be between two funding ecosystems with different definitions of “safety,” “innovation,” and “public interest.”
The Bores race illustrates how those definitions are contested.
Supporters of guardrails argue that frontier AI systems pose risks that cannot be managed through voluntary best practices alone. They want enforceable requirements, oversight mechanisms, and a clear standard for what companies must do before deploying powerful models.
Opponents often argue that heavy-handed regulation can stifle progress, push innovation overseas, or create compliance burdens that don’t actually reduce risk. They may also argue that safety should be handled through industry-led frameworks or through a slower, more consensus-driven regulatory process.
In the Bores campaign, those arguments were translated into political messaging. The RAISE Act became the proxy for one side’s vision of safety governance. The super PAC’s involvement became the proxy for the other side’s belief that the law went too far—or moved too fast.
But the most revealing part of the story is what happened to Bores himself.
He was not merely attacked; he was spotlighted. His legislative work became the centerpiece of a national conversation. That kind of spotlight can be a gift and a curse. It can elevate a candidate’s credibility with some voters, but it can also make them vulnerable to being defined by a single issue—especially one as emotionally charged and technically complex as frontier AI.
When a candidate is reduced to one policy fight, they lose the ability to demonstrate competence across the full range of issues voters care about. Even if voters like the candidate, they may still choose someone else who feels more aligned with the district’s broader priorities.
That dynamic is common in modern politics, but it’s amplified when the issue is both high-stakes and difficult to evaluate quickly. AI policy is not something most voters can assess in a weekend. They rely on cues: endorsements,