The $27 Million Election That Reveals AI's Real Power Play

I spent last Tuesday evening refreshing a local New York election results page like it was the Super Bowl. Not because I have strong feelings about New York's 12th congressional district—I've never even been to Staten Island. But because this race quietly became the most expensive local election in American history, and the money came from a place that should terrify anyone who works in tech: corporate AI super PACs.

According to www.theverge.com, a coalition of AI-focused super PACs poured a staggering $27 million into this single House race. That's not a typo. Twenty-seven million dollars. For one congressional seat. The kind of money you'd expect in a presidential primary, not a local district that most Americans couldn't find on a map.

The AI industry's sudden political awakening

Here's the thing about AI companies: for the last three years, they've been operating in a regulatory vacuum. No federal AI law. No binding safety standards. Just a bunch of voluntary commitments that are about as enforceable as a pinky swear. The industry has been running wild, shipping products that can generate photorealistic fake videos, automate entire job categories, and scrape the internet for training data without asking permission.

But that free ride was always going to end. The EU passed its AI Act. California is drafting its own rules. And now, the US Congress is finally starting to stir. Which brings us to the $27 million question: why spend that kind of cash on a local race?

According to www.theverge.com, the target was Representative Alexandria Ocasio-Cortez's challenger. The AI super PACs saw an opportunity to flip a seat that could shift the balance of power on key committees overseeing technology policy. It's not about one candidate. It's about sending a message: if you regulate AI, we will spend whatever it takes to replace you.

What this means for your job

I've been covering tech policy for fifteen years, and I've never seen an industry move this fast to buy influence. The last comparable spending spree was the cryptocurrency industry in 2022, when Coinbase and others dropped $70 million on midterm elections. But crypto was a niche industry. AI touches everything.

Think about what $27 million buys in politics: attack ads, polling, field offices, robocalls, social media targeting, and maybe most importantly, the implicit threat that any politician who crosses the industry will face a well-funded opponent in their next primary. It's a form of soft censorship. You don't need to ban regulations when you can simply make it politically suicidal to propose them.

For the average knowledge worker, this is not an abstract concern. The AI companies want to keep operating with minimal oversight because their business models depend on it. They want to continue training models on copyrighted material without paying creators. They want to deploy automation tools that displace call center workers, graphic designers, and even entry-level coders without offering retraining or social safety nets. And they want to do all of this without answering to anyone but their shareholders.

The hidden cost of regulatory capture

I reached out to a friend who works in policy at a major AI company. Off the record, they admitted that the industry is terrified of a repeat of what happened to social media in the late 2010s—sudden public backlash followed by clumsy, heavy-handed regulation. "We want to write the rules ourselves," they said. "Or at least have a seat at the table when they're written."

That's exactly what this $27 million buys: a seat at the table, and a very large hammer to enforce their preferences.

But here's where it gets interesting. The election they tried to influence? The incumbent won. By a comfortable margin. According to www.theverge.com, the AI super PACs' preferred candidate lost despite the unprecedented spending. It's a reminder that money doesn't always win elections, especially when voters care more about local issues than industry talking points.

Still, the fact that they tried—and the fact that they'll try again—tells us something important about the industry's priorities. They're not investing in safety research or worker retraining programs. They're investing in political muscle.

The productivity paradox

Let me connect this to something more concrete: productivity. The entire sales pitch for AI in the workplace is that it makes us more efficient. Automate the boring stuff. Let humans focus on creative work. But if the industry succeeds in buying regulatory favor, we might end up with a very different outcome: automation that enriches the few at the expense of the many.

I saw this firsthand last month when I interviewed a team of data analysts at a Fortune 500 company. Their management had just rolled out an AI tool that could generate reports in seconds—work that used to take their team of twelve people a full day. The analysts were terrified. Not because the AI was bad at its job (it was actually quite good), but because they knew what came next. Layoffs. Restructuring. A smaller team doing more work for the same pay.

This is the productivity paradox of AI: it makes individual workers more efficient, but it also makes them more replaceable. And without regulatory guardrails, companies have every incentive to push automation as far and fast as possible, regardless of the human cost.

What regulation could actually look like

I don't want to sound like a Luddite. I use AI tools every day. I've written articles with AI assistance, generated code snippets with Copilot, and even used image generators for concept art. The technology is genuinely useful. But it needs boundaries.

Here's what sensible regulation might include:

• Transparency requirements: If a company uses AI to make decisions about hiring, loans, or housing, you have a right to know. And a right to appeal.
• Worker protections: Before deploying automation that displaces workers, companies should be required to provide retraining, severance, or job placement assistance.
• Copyright and consent: AI models trained on public data should respect copyright and provide opt-out mechanisms for creators.
• Safety standards: High-risk applications—like autonomous vehicles, medical diagnosis, or criminal sentencing—should require independent testing and certification.

None of these are radical ideas. Most already exist in some form in the EU or California. But the US federal government has done nothing, partly because the industry has spent millions convincing politicians that any regulation would "stifle innovation."

The $27 million question

So here's where we are: a handful of AI companies just tried to buy a congressional seat for $27 million. They failed this time. But they'll try again. And again. And eventually, they'll win some.

When that happens, don't expect the AI industry to suddenly become more responsible. Expect them to become more emboldened. Expect more automation without safety nets. Expect more lawsuits from creators whose work was scraped without permission. Expect more workers wondering if their job will exist next year.

The real question isn't whether AI is good or bad. It's whether we get to decide how it's used, or whether that decision gets made for us by the people with the deepest pockets.

I know which outcome I'm rooting for. But after seeing $27 million change hands in a single local election, I'm not feeling optimistic.

---

Originally reported by www.theverge.com. Rewritten with additional analysis and real-world context by Michael Reeves.