Why is the AI ​​industry scared of this Palantir alum running for congress?

As the midterm election primaries inch closer, some candidates are focusing their campaigns on how they’ll regulate artificial intelligence. On the right, populist Republicans are warning that the AI ​​industry stands to undermine the Make America Great Again movement. On the left, there’s worry about the sector’s growing political and social power. Across the spectrum, there’s near-universal concern about what the technology might be doing to children.

The donor class is now getting involved: A super PAC called Leading the Future backed by OpenAI executive Greg Brockman and Andreessen Horowitz plans to spend as much as $100 million in the midterms to support its preferred candidates. Another bipartisan super PAC, focused on pushing for a national framework on regulating AI, formed earlier this week. These fights come as the Trump administration pushes to limit the ability of states to regulate the technology.

Alex Bores, who authored legislation on AI in New York state and is running to represent its 12th district, has become an early target for Leading the Future’s political spending during the midterms. “It’s a badge of honor,” he says, comparing the effort to an F rating from the National Rifle Association.

“This is not tech versus everyone else,” he tells. Fast Company“This is one small subset of the tech ecosystem that, instead of engaging in collaborative discussions on bills and how we can work for all, has decided they want to drown out the voices of anyone who isn’t them by spending hundreds of millions,”

Fast Company chatted with Bores about AI in politics, his time working at Palantir, and what it might take to modernize the government. This interview has been edited for length and clarity.

One of the things the draft executive order talks about is creating a federal approach to regulating AI by the Trump administration. How seriously do you take that?

It’s a cliché in DC that when you want something to not get done, make a commission to study it. And so making a proposal to study a thing—to maybe put a policy forward—is silly. I want to be clear, the correct answer to these questions is a federal standard. The only reason the states have been acting is because the federal government hasn’t. If they want to actually work on a federal standard, they will find partners across the aisle. What they are prioritizing is stopping any state from taking action, not actually solving problems.

We’re in a moment when there’s a lot of criticism and debate about what the future of the Democratic Party should look like. In New York, a liberal stronghold, I’m curious about where conversations about how we should handle emerging technologies and AI might sit in the remaking of the Democratic Party—and the push to focus on issues that speak to younger voters and disaffected voters who haven’t been so into what the Democrats have been offering.

We should always be human first and human focused. The specific ways that plays out in AI policy are ones that speak to younger people. The biggest impact of AI on the job market right now is on entry-level jobs, and you’re seeing a rise in unemployment of people in that cohort looking for their first job.

One of the most popular things we did in New York this year was phone-free schools and making it so that we could actually change how tech is used in the classroom and make sure that it’s been used for education and not for screening or scrolling on social media. ,Earlier this year, the state instituted new rules on electronic devices during school hours.,

We shouldn’t shoot for one big grand bargain on AI, as if it’s a static issue. It is something that infuses everything we do, and we need to continually be updating our protections as the technology grows.

Can you talk a little bit more about what you did at Palantir and your decision to leave?

I was at Palantir for four and a half years, I spent the vast majority of that time on the federal civilian team, and so it was working with the government to better serve the American people, I worked with the Department of Justice to go after the opioid epidemic, , , , I worked with the Department of Veterans Affairs on better staffing their hospitals and better serving veterans, I worked with the CDC (Centers for Disease Control and Prevention) on understanding epidemics,

I’m really proud of the work that I did there, because it was all about actually making government work better. Separate from tech, I think that’s the thing that the Democratic party and the country as a whole needs to see more of, is people who are willing to push through the obstacles and make sure that government is actually a force for good and serving people, and not just the political mudslinging that is most of what they see on TV.

I left in 2019 when Palantir renewed—or soon after—Palantir renewed their contract with ICE, Palantir had a contract with Homeland Security Investigations to help with fighting cross-border drug trafficking and human trafficking, During the first Trump administration, they started using that software for Enforcement and Removal Operations—for what most people think of as deportation, That’s a different department within ICE, , , , That wasn’t something that was foreseen when the contract was signed in the Obama administration, And when Palantir renewed that contract, without cutting off that work or putting in protections that would step up in the future, that motivated me to leave,

What do you make of the conversation surrounding Palantir right now? The company has insisted that it’s worked through multiple administrations, but the work with ICE, as well as with Israel, has sparked major criticisms of the company.

I haven’t been there for six years, so I don’t have more detail on how they currently operate than anyone reading the news, but I’m proud of the work that I did there and very public with the reasons why I left.

How hard is it to buy technology in government to make things faster, more efficient? Government modernization—updating government software and providing better customer service—continues to be a big challenge.

No one asks about government acquisition. This is amazing. It’s horribly inefficient, and that hurts the American people. It takes the government far too long to be able to sign a contract, so the cost—therefore what the American people end up paying—rises.

It then benefits sort of the insiders who know how to do contracting more than the people who can deliver useful services, so then the American people don’t get the benefits from startups and others that might have cheaper, faster ways of solving problems.

It is a problem at every level of government, One of the first bills I passed in New York, as an Assembly member, was making it easier for the government to actually use cloud computing instead of buying servers and always running on hardware, which slowed down getting services to New Yorkers, , , , A program called FedRAMP was supposed to make it easier to get tech to the government: You would do one security screening, be certified, and then be able to sell to each (federal) department, so they didn’t have to do their own (screening), But it has become an incredibly onerous process that just makes it much more difficult to actually work with the federal government,

Should politicians be using generative AI in advertisements?

I haven’t done it. But I have written the laws in New York that regulate it, and what we came to was it has to be disclosed. And if you use it to make deepfakes of a candidate, the candidate has the right to sue for injunctive relief with expedited review to pull that ad down if it’s going to deceive the public into something that actually happened.

The problem of deepfakes is one that has a technical solution, and policymakers just haven’t kept up, Historically, we’re told it’s, , , just going to be a cat and mouse game where you have better detectors of deepfakes and then better AI generators (or that) we’ll never win that battle, But the industry has created a free, open-source metadata standards of data called C2PA that can be attached to any standard audio, video, or image file type that cryptographically proves whether that piece of content was taken from an actual device, was generated by AI, and/or how it’s been edited throughout the process,

If you got to a place where 90%, 95% of people were using that standard, we’ve solved the problem of deepfakes, because anytime you don’t see that credential, you would immediately be suspicious of what’s being shown.