The ‘Political Disaster’ Looming for Democrats on AI

By now, the string of alarming recent headlines blur together: an Anthropic engineer who quit out of fear that his company was downplaying the possibly humanity-ending risks of its product. Swarms of rogue AI agents attempting to deceive their human overseers. A string of dire warnings from AI executives and luminaries. Such apocalyptic talk has been a regular feature of the tech world in recent years, but the rapid-fire developments of recent weeks have put it on the front burner for everyone else, most notably in Washington. Until just the last few days, Congress — where AI knowledge among its oft-geriatric members tends not to be expansive — had mostly been twiddling its thumbs on artificial intelligence.

That’s where Rob Flaherty saw an opening. Flaherty, who was the Deputy Campaign Manager for Joe Biden and then Kamala Harris in 2024 (and whom I spoke with earlier this year about the infamous DNC autopsy), recently co-founded a group, Tomorrow Together, which will fund and advise Democratic candidates around AI in the midterm elections and beyond. Flaherty believes that artificial intelligence will be the defining political issue of the next several years, and that Democrats need a far better and more unified message than they have presented so far. His efforts have coincided with a moment where everyone in politics is suddenly paying close attention. I spoke with him about what both parties are failing to understand about AI, what sensible regulations might actually look like, and why President Trump is so politically toxic on this subject.

You recently told Politico that on AI, “Voters don’t know which party is better. They want solutions that look a lot like things that Democrats want. They still don’t know whether a Democrat or a Republican is going to be more helpful.” You went on to say that there’s a huge political advantage in taking the issue seriously now. I’m wondering how much of this initiative you’re leading has to do with your personal level of fear and apprehension around the technology, and how much is that you see it as a rare kind of political opportunity.The truth is that it’s both. I am super freaked by what this technology can do, and that makes me completely in line with the average American. People are actually concerned about this. This is one of those opportunities where good policy is good politics, and doing the things that we need to do to step up and regulate these companies, provide a really strong set of protections and guardrails, build a society that can exist in the context of powerful AI in an economy that doesn’t concentrate the wealth with six rich guys. That’s important for continuing democracy and making the country work, but also a thing that a political party would have a lot of advantage by jumping on.

And as you said, people don’t really know who’s advocating for what. If the average voter could name anyone’s political position on AI right now it would probably be Trump, who is loudly against any regulation and totally blase about the risks.It is one of the truly most bafflingly unpopular political positions he could be taking right now.

As Matt Yglesias put it, Trump is below the“legalize heroin” line in public opinion on this..Exactly. Trump is historically unpopular and maybe he just doesn’t care, but I think he is dragging the Republicans to a really bad place here. It’s one thing to be holding the bag when a bad thing happens. But if a bad thing happens, the worst possible thing is to be the person who said it wasn’t going to happen in the first place. So voters are screaming at Washington saying they need to take this seriously. If you look at Mike Rogers, who is running in a competitive Senate seat in Michigan, he’s calling for pacing the frontier and regulating the AI companies. That is more of an indicator of where Republicans who actually have to deal with voters are.

Trump does seem indifferent to public opinion at this point, and he’s doing a lot of unpopular things, like continuing the Iran War. But why do you think he is so all in on AI? Is it a financial motive? Self interest? I can’t quite glean what’s happening there.Who knows how he’s making money? What I will say is the reason Donald Trump is here is the reason Donald Trump is Donald Trump. You can see this in the reporting. There are a number of people in his White House who are incredibly concerned about what this technology can do, going all the way up to Susie Wiles. But at the end of the day, a bunch of rich guys can call Donald Trump and get access. And the guys who are most financially incentivized to have no regulations or oversight for their tools are dialling him up and telling him, “Don’t do anything here.” Mark Zuckerberg, Jensen Huang, David Sacks — all these guys are in his ear.

So it’s Donald Trump — a rat king of self-motivated profit — and a bunch of rich dudes trying to make it so this industry continues to concentrate wealth in their hands. I think the precise thing that is happening in this moment is the precise thing that leads him to do things that are horrifically unpopular with real people.

You’ve said that you want the group you’ve founded to be a clearinghouse of AI ideas, not a place for you to advocate a specific stance. But what are some basic rules that you think both make sense and are politically possible to execute?With Tomorrow Together, we started from the position that this is going to be the biggest issue in the next election, that Democrats need to be on top of it, and that we need to get people to pay attention. For a long time before we launched, our plan was to come in and try to make Democrats care about this thing as it rises in salience. It has now risen in salience.

Yeah, the last two or three weeks have been something.We launched and Jacob Coxon tweeted the same night. Obviously we had nothing to do with that, but it was fortuitous timing. So now we’re in a position where voters care about it. They don’t know who’s better on it. And elected officials are like, “I don’t know what to do.”

We don’t really have a lens on the ideal policy. But I think there’s a couple of things that are pretty clear. Voters want to regulate these companies. That is pretty straight up: they believe the government has a role in making sure that these companies can’t just do what they want. Do they necessarily trust that the government’s going to do a good job? No, but they believe that it is the government’s role, the way in which there’s a system by which they get agency over it. So whether that looks like independent mandatory evaluations — that feels like a pretty obvious floor. Does it feel like kill switches? Who knows if that’s even possible, but to the extent that it is, great. What I think is a non-starter is anything with the word “voluntary” in it.

You don’t think Skynet should police itself? Sorry, I had to get the Terminator in there somewhere. Exactly. Where has that gone wrong before? It’s so funny to see Mark Zuckerberg be like, “Well, the markets incentivize us to have safe products.” And I’m thinking, “Brother, you’re Mark Zuckerberg.”

But I do think there’s this other thing, which is that there’s a bigger sense of control that voters want over this. AI is entering the political lexicon against the backdrop of people feeling like there are all these forces that are making their lives worse. They can’t afford shit. They feel like everything is optimized for profit. They feel like a bunch of guys are getting rich and it ain’t them. And then you factor in this technology where a bunch of dudes are like, “It could either take your job or kill you.” They’re super mad about that. When they think about AI risks, voters are thinking yes, it could blow up the world, but they’re also thinking, is it going to come for my livelihood?

So even if your P(doom) is 10 percent, you still think there’s a 90 percent chance that humanity is going to continue, so you have to also have a plan for what that world looks like. To many people, that will look like taking on these companies, taxing them, but also providing a set of economic policies that make sure that they benefit from whatever this thing is going to be. You see this in the polling: huge margins for portable benefits, huge margins for job guarantees for people who get laid off as a result of AI. There’s a real question of how we want to build a society around a world where powerful AI exists, which I think is really important in addition to getting the safety stuff right.

This just sounds like such an enormous challenge for Congress, especially given the fact that they’re just now becoming AI-literate.If they even know what it is at all. But one thing here is that this is really the first time voters have stood up for tech regulation as a general thing. There is a huge mandate for this. And there’s a huge mandate for policies that leverage the government to serve as the voice of the people and regulate and protect people from the worst possible outcomes.

America has been distinguished by a lack of regulation on any big tech stuff. It’s basically been nothing for decades.Nothing. And that’s because voters have ultimately been like, “Yeah, this kind of sucks, but it’s not a high priority.” This is now becoming a high priority. And there’s also interesting coalitions forming. I think Republicans need to feel pressure on this,, but the incentive structures are different here because they are so far on the wrong side of voters, like 80/20 on the wrong side of voters. That gives me a general sense of hope that something is achievable here. And the other thing is that ultimately, provided the AIs don’t release a rogue anthrax that kills millions of people, is that there’s going to be a race to the top on this, at least on the Democratic side, and an incentive to take real action as a key thing that Democrats are going to do if they win.

So I think you’re going to see interesting fissures on the Republican side as well. where Ron DeSantis and Josh Hawley and Ted Cruz all have different positions than JD Vance in the White House. And I think you’re going to see this discussion toward a regulatory structure on both sides.

Do you really trust the likes of Josh Hawley on this stuff? Some of these Republicans like to talk a populist game and then when it comes down to it, are they really going to vote to strictly regulate large corporations, which is generally not in their blood?There was the big Pro-Human summit and you had Bernie Sanders being like, “We have to cooperate with China.” And then the next speaker, Steve Bannon, is like, “We have to deport all the Chinese.”

I think the coalition of Republicans looks a lot like Bannonites, religious folks who hate the idea of a machine God, and frontliners who are afraid of voters. And you can start to triangulate what their political incentives look like. So, do I trust Josh Hawley? No. Do I believe that there are areas of consensus that we will be able to find with Republicans on this stuff? Absolutely. And that should be encouraged. I mean, at the end of the day, there’s a real shot that even if powerful AI and super intelligence does not eradicate humanity, it makes the tenets of a basic democracy really, really hard. So my hope springs eternal that Republicans will move to the right place on this. And if not, they should be held accountable.

Another possibility here is that this is all overblown, that the people sounding these alarms are overrating the chances of catastrophic damage from AI. And the alarm has ramped up so quickly that you could see Congress being rash because they feel they have to do something. This isn’t the best analogue, but I’m thinking of legislation like the Patriot Act, where it was like, “Well, we’ve got to address terrorism now,” and it was a rushed, overreaching piece of legislation.We just don’t really have time. The models are going to get increasingly capable, even if you don’t buy all the theories. The skeptical left is just trying to hold on. They’re wrong, they know it, and they just have to find any evidence that they’re right. The reality is even if you just buy that it is a bad thing that technology should be able to figure out how to make itself better and better without humans watching it, such that who knows what it does, stopping or slowing or managing that process is a good thing.

We are in a race with China — that is absolutely true — but their incentives are for this not to destroy the world too. So I think in creating the environment and the structures for safe AI, we have to take some swings here. I would rather us go big and adjust than underdo it and have a catastrophic incident or mass job displacement or whatever comes from that.

Speaking of strange bedfellows: the Bluesky “AI can’t do anything right” crowd is weirdly in line with the Trumpy “don’t even try to regulate AI” crowd, for completely different reasons. They both think the doomsayers shouldn’t be trusted. Yeah. A bunch of people have been sold the idea that nobody wants this and it doesn’t work, and that is just not true. Consumers are really skeptical of it; they don’t like what it means. But boy, enterprise loves it. And that’s where a lot of this will go down. I think there are people whose lens on this is a chatbot that sometimes fucks up versus a model that exists with a different interface that can make itself better and better and better and better, and over the course of a couple years is going to make it so good that it can replace your job. That should be frightening. And there is this reflexive set of priors that leads people to not see what’s in front of them. On the left, it is so fucking infuriating because if you’re somebody who cares about universal benefits, taxing the wealthy, all this stuff, voters are looking at AI and going, “Let’s do those things.” This is the best moment ever for the Bluesky, left crowd. But to make people see that — there’s just some people who are so stuck in their priors. And again, that’s a gift to David Sacks.

If you’re looking at whose jobs are going to go first from this thing, it is college-educated, early-to-mid-career people who largely live in urban environments and largely vote for Democrats. When those people start to become economically disaffected, that is a blinking red coalitional problem for Democrats. And the conditions that create are a lot of room for economically populist, anti-tech, anti-immigrant racist. It creates the conditions for Tucker Carlsonism if we don’t get ahead of it.

There is a political disaster that is around the corner for Democrats if we don’t take this seriously and if we don’t push here, and there’s a political disaster for Republicans if they don’t act seriously here. And it’s invisible to both sides, and so you see versions of each side making tremendous, I think, generational mistakes. Luckily, I think Democrats have started to move on this and some Republicans, but again, we need everybody to really figure this out.

How much tangible progress have you seen from Democrats in the last couple weeks?You are seeing people take AI seriously for the first time and also realize voters care about it. On a policy side, you’re seeing people make this loudly and proudly the center of what they’re about. You have Josh Shapiro, who was like “Let’s do data centers wherever, guys.” Now he’s calling for strong regulation of the tech companies, and he’s calling for a data center moratorium. The politics are moving. I’m sort of squishy on a data center moratorium — It’s where voters are, but it’s not where the issue’s going to be — but you’re seeing people move.

I think there’s two major problems here. One, there was that article this week that said the Democrats aren’t going to talk about AI. They’re going to talk about the cost of living. And that to me is just a complete and utter fucking miss of how people are grappling with AI, which is against the backdrop of their economic situation. It’s like, “Oh my God, I can’t afford shit. And now this thing is going to come and take my job. Are you kidding me? This thing is going to come and poison my kid’s brain. Are you kidding me? These three dudes are going to get rich off this thing. Are you kidding me?” And so to me, to just be like, “We’re going to talk about groceries,” just completely misreads what this issue is about.

And the other thing that I will say, and I am experiencing this a lot: the only real people who are in many of our elected officials’ ears about this technology are the frontier labs themselves, and that’s a real problem. OpenAI’s lobbying team is not a good actor here, as we’ve seen through their electoral spending against regulations. So there is this fundamental problem where our officials don’t understand it. They don’t understand the tech, and the only people they’re talking to have a vested interest in making money off this thing.

I was talking to a colleague about what it would have been like if AI had gotten big in 2015, when Obama was president. His message might have been a sunny “Let’s harness AI for good,” mixed with warnings about what could go wrong. That’s how he’s talking about the issue now. Should Democrats try to maintain an optimistic tone about the possibilities of well-regulated AI, or should they run with the Republican playbook of scaring people? How do you view that balance?I think if we just left this that we’re going to scare people in order to win, we would miss a massive opportunity here. I associate with what President Obama said — I’m neither accelerationist nor doomer. I think there are real society global level dangers when it comes to what this technology could do, and we should not be accelerating. We should be pacing, not just because of safety issues. The economy is not ready. We don’t have the structures to protect people from this thing at all. But I’m also not a doomer.

I think there’s probably a chance it kills everybody. The people who are working on it say that. Look, it makes sense. Any model that can build itself to a certain level of hacking strength and take over a computer — anything that can kill people with a computer is at risk. So we have to protect that stuff. It’s a narrow window, but there’s a world where this technology leads to broad shared prosperity, better lifestyles, and incentivizes the transformation into a fairer new economic bargain between the people and corporations and the government. I really genuinely believe that that world could be around the corner, but it is not currently headed there. And we are currently headed towards the worst-case scenario where Sam Altman, Dario Amodei, Jensen Huang, Mark Zuckerberg, Sergey Brin…

You’re forgetting Elon, the best of them all. …and of course, Elon Musk get fucking loaded off of this while everybody else becomes a feudal serf. And by the way, this thing maybe just decides to take out California’s power grid. I think that’s where this is moving. I think we have to, as a country and as a society, adjust that direction of travel. And whoever does that, whether they are Republicans or Democrats, will win the political moment of the next decade. I am a Democrat. I want it to be Democrats, but I hope everyone is competing over how best to do this.

This interview has been edited for length and clarity.