Why New AI Super PACs Plan to Spend Millions in the 2026 Midterm Elections

In just the past week, my Spanish teacher suggested our class use AI to generate tailored lessons; a friend sent me AI-generated music he made; and another friend told me about an AI research tool that’s saving her time. This is just how it is now, it seems. AI will keep becoming an ever-bigger part of our daily lives.

But it’s not inevitable that we sit and watch as tech companies develop AI platforms on their own terms, remake our society, keep us addicted to fossil fuels, and continue to amass economic and political power while doing so. In recent weeks, tech companies have launched multiple new PACs, so I spoke with Sacha Haworth, executive director of The Tech Oversight Project about how these new war chests will be deployed to derail efforts to regulate AI–and defeat good candidates for office.

“We cannot trust that these companies whose sole mission is to make money will regulate themselves. They never have and they never will,” Haworth told me. While many tech CEOs have testified at Congressional hearings, making overtures about how they’re going to work to uphold safety measures and democracy, Haworth says: “Believe what they do, not what they say. Because those same tech CEOs privately lobby against all regulation, all oversight. They’ve dismantled the same teams that they assembled under pressure from Congress to make sure that kids aren’t being matched with pedophiles online. If there aren’t regulation guardrails in place, there’s nothing to stop them from dismantling safety measures again.”

Meta has run a federal PAC since 2011. This August, the company started another super PAC, Meta California, which will try to force out of office lawmakers perceived as not friendly enough to AI. Then, in late September, Meta announced the super PAC American Technology Excellence Project, which will spend “tens of millions of dollars” electing both Democratic and Republican state-level candidates. This PAC could do outsize damage because state legislative races are relatively inexpensive investments for major donors to influence, and because states are a venue where tech regulating legislation is more likely to pass given the GOP’s grip on Congress. According to Axios, 1,100 state-level tech regulation bills have been introduced just this year.

A pair of billionaires, venture capital firm Andreessen Horowitz, and Greg Brockman, co-founder of OpenAI, have helped start another super PAC, Leading the Future, which also aims to curtail efforts to regulate AI, by chipping in $50 million apiece.

Share

To really understand the gravity of those billionaires spending all those millions to let their businesses run free, we have to first understand the real cost of using AI. Although the technology does have positive uses, its harmful consequences are dire and wide-ranging. Some of the ways that AI can be harmful:

• AI consumes obscene amounts of electricity, the production of which currently relies on fossil fuels and coal. According to the MIT Technology Review, citing numbers from Lawrence Berkeley National Laboratory: “By 2028, more than half of the electricity going to data centers will be used by AI. At that point, AI alone could consume as much electricity annually as 22% of US households.” Experts anticipate that those data centers will use dirtier forms of energy, including coal, at a moment when there is a moral and pragmatic global imperative to reduce emissions to maintain a liveable planet.

• AI raises our bills. Electricity rate payers are footing the bill for all that additional energy consumption that AI companies are using to generate profits for themselves. A Bloomberg report from last week details how “Wholesale electricity costs as much as 267% more than it did five years ago in areas near consumers. That’s being passed on to customers.”

• AI perpetuates discrimination. AI technologies are being used to make decisions that affect people’s lives in major ways. They’ve also been found to make discriminatory decisions that replicate long-standing injustices. Banks are using AI to approve or deny mortgage applications, for example, and they’ve been found to discriminate against people of color. In 2021, The Markup found that 80% of Black applicants, 70% of Native American applicants, and 40% of Latino applicants were more likely than white applicants with similar financial statuses to be denied a mortgage by AI. Recent academic research has backed up that AI making mortgage decisions discriminates.

• AI displaces jobs. Generative AI is already replacing human employees in certain fields, and there’s an expectation that it’ll displace lots more workers. There’s no real way to know how profoundly developments in AI will change entire industries, job markets, and millions of people’s ability to find work and make a living.

• AI distorts our sense of reality. We’ve all seen the deepfakes traveling on social media. This type of misinformation further alienates us from the truth and drives even greater wedges between people who need to unite if we’re going to turn this country around.

• AI harms children’s mental health and endangers their lives. Teenagers have died by suicide after AI chatbots apparently encouraged their suicidal ideation. A study out of Stanford showed that it was easy to elicit responses about self-harm and violence, among other sensitive topics, from chat bots, and the study concluded that the platforms posed specific risks to adolescents. Tech companies have lobbied against California legislation designed to make platforms safer for kids and adolescents.

• AI potentially changes life as we know it. No one really knows where all this is headed. Tech safety experts Eliezer Yudkowsky and Nate Soares recently published a book titled, If Anyone Builds It, Everyone Dies wherein they argue that super-intelligent AI is going to end life on this planet sooner than later. I hope they’re very wrong, and I also believe I’m not equipped to evaluate their arguments.

Tech company lobbyists and political strategists realize that we’re aware that AI poses risks. “They know they can’t change public opinion. Americans – left, right, old, young – are universally united in seeing the harms that AI might have and the harms that social media already has,” Haworth told me. “They want their elected officials to do something about it.”

That means that so-called “AI-friendly” PACs probably aren’t going to run ads or push talking points related to their issues. “They’re not going to run ads in your district saying AI chat bots are good for kids,” Haworth says. They know no one is buying that. “No, they’re going to run negative ads attacking [their preferred candidate’s] opponent about something else. They’re going to run ads with whatever lies they’ve come up with. So, I wouldn’t be surprised if you see tech money funding some really ugly smears against Democrats in upcoming elections.”

The new AI PACs will likely follow the playbook that cryptocurrency industry-backed PACs used to beat former Ohio Senator Sherrod Brown a cryptocurrency titan, Bernie Moreno, ran against him in 2024. That race attracted more ad spending than any other in Senate history, according to CNBC, and the crypto industry spent at least $40 million to defeat Brown, who had voted to impose sensible regulations on the crypto industry (which is currently operating with very few guardrails, meaning everyday investors have tended to lose big while those capable of pulling strings, like President Trump and his family, have further enriched themselves).

Meta, along with other tech companies, will likely play a major role in the 2026 midterms. When describing this, Haworth emphasizes an important point: Because Mark Zuckerberg has a unique position as owner of Meta, the new Meta super PACs give him–one individual–a vehicle for deploying enormous amounts of money to shape elections and policy discussions according to his personal wishes, without the kinds of disclosure requirements he’d have to meet if he were giving as an individual. That’s a lot of power for any one person.