Ellen Schwartz Grapples with AI’s Power and Guardrails
In June, a new US Super PAC launched with about $5 million and a simple goal: make political candidates who accept money from Frontier AI interests… costly. Its main opponent, Leading the Future, claims to have raised more than $140 million, from backers including OpenAI president Greg Brockman and Andreessen Horowitz.
Ellen Schwartz is COO of that scrappy super PAC, Guardrails Alliance, and its sister organization, Guardrails Action. We talked about why citizens can’t truly opt out of AI, what the industry really means by “pro-innovation,” what money can’t buy, and other riffs on “Big AI” in politics.
We also asked her a question that’s uncomfortable for us at AlignIQ. We can have a business model basically because organizations feel pressure to adopt AI, so are we part of the problem?
Guardrails is a US political organization, and AlignIQ isn’t affiliated with it. We’re sharing Ellen’s perspective to demonstrate the ripple effects of who looks out for us when, collectively, we’re uncritically accepting AI into our workspaces and without option, our lives.
Answers have been edited for length and clarity.
Caroline: Please define what you do, maybe in terms of the fact that what you’re doing is both with a PAC and a 501c. So what does each of those things encompass for you?
Ellen: I am the chief operating officer working across two sister organizations that are independent organizations from one another: Guardrails Alliance, which is a super PAC, which can raise unlimited contributions to spend independently supporting or opposing candidates in elections, and then Guardrails Action, which is a 501(c)(4), which is similar to a 501(c)(3) nonprofit organization in that it’s tax-exempt, but it can do unlimited lobbying toward its mission, to build the political power and conditions necessary to pass regulation on AI and to work towards a transformative economic agenda for AI for the public.
We launched in mid-June, with the purpose of making big industry pushes for deregulation of AI politically toxic. Leading the Future (LTF) had come in with the intention to spend a ton of money to electioneer a deregulatory agenda. It’s supported by AI executives and venture capital executives. I think between that and the sort of aggressive authoritarian push within the tech industry and tech leaders, it became clear that there’s a real danger to our democracy, and AI is acting as an accelerationist force for all the things that put at risk everything we care about. I think Guardrails Alliance wanted to make that money toxic early, so we’ve been following LTF into races where they are spending or endorsing candidates, to basically counter that. Get candidates to disavow support of LTF, creating political cost and accountability for taking that money, that kind of thing. We recently did a small amount of work in the 2026 primaries and expect to do some more spending as we near the U.S. Midterms.
Guardrails Action is doing a significant amount of organizing work, both on Capitol Hill and with allied organizations. And I would say when we think about our coalition, we think about the progressives, the “tech left”, labor, and the AI safety sector, which has become a major political player recently, and where folks have been giving a lot of political money, especially to counter groups like LTF.
Caroline: You have the support of a lot of people who work in AI, I imagine.
Ellen: Those who have made their money in tech. But that also includes tech employees–OpenAI employees–who were angry about Brockman giving $25 million to LTF.
So right now we’re focused on that, and especially looking ahead towards 2028, where we think AI will be a big issue where we need a unifying rally cry.
Caroline: What would winning look like for you?
Ellen: I think the dream would be having political majorities able to pass legislation that is both publicly very popular and politically safe: we’ve created the conditions where politicians could take bold stances on AI that benefit the public and challenge corporate power, oligarchy, monopoly, and be empowered to do so, and where we have a political environment where regulation can pass and create transformative change. There is political realignment happening across our political spectrum. And there are Republican groups that are doing work on this as well, especially with data centers and surveillance. We are pretty focused on the Democratic side because that’s where our network and connections sit. The way the AI corporations are functioning is unbelievably unpopular right now.
Caroline: That anger should be harnessed. There’s also a little bit of a wider prisoner’s dilemma of AI, of people feeling like they have to use AI because everybody else does. How would you say that lobbying is going to help get out of that trap?
Ellen: Well, I think right now we actually don’t really have the ability to opt out. Yes, you as an individual might be able to choose whether or not to use Chat GPT or Claude, but we don’t actually have the ability, at a systemic level, to opt out. AI is being used on us. AI is used in decision-making, it’s displacing and degrading work… I think lobbying can help.
The right legislation would allow for an individual to have control over the way that they are engaging with AI. If you’re a senior, you’re on Medicare, you can’t control whether or not an AI system is being used to make decisions on your benefits eligibility. If you are shopping, there has been [evidence of price] surveillance of your buying habits. You don’t have the ability to opt out. We should be able to choose!
There should be domains where AI is just not acceptable, and we should have a shared set of principles and standards, societal agreements about how this technology is treated, from the development stage to before rollout to addressing harms after the fact. Right now we don’t have those things–we don’t even have a system to hold thing [and people] accountable.
It goes beyond an individual choice right now. I think of it similar to climate change: what I do or do not do at an individual level is not going to affect its impacts at large. There needs to be a collective addressing of harm and impact that is coordinated, right, standards set for governance.
A lot of the people who work on the forefront of this issue are passionate about technology and want to see it work for people, but what we have right now is not the right service delivery model; it’s not fit for purpose. Profit as a sole motive is not, and scale for the sake of scale is not how we should define success.
The bar is in Hell right now. We’re not enforcing the basic laws on the books.
Ellen schwartz
Caroline: I was looking at LTF’s activities and I noticed two things. They say that they back pro-innovation candidates, which might imply that you aren’t. So how do you answer that?
Ellen: I would say that’s kind of the only way they can frame themselves positively. But I think people see through that pretty clearly. There’s always going to be some tension between innovation and…
Caroline: Regulation.
Ellen: …but the bar is in Hell right now. We’re not enforcing the basic laws on the books. And what they’re trying to do is make AI above the law. I think the overwhelming majority of Americans agree AI should not be above these well-worn paths and laws that are in place for a reason. And it’s not exceptional enough for those principles not to matter.
Caroline: The other way I saw how they characterize themselves is that they’re trying to beat China. So what do you hear with that?
Ellen: I think in general those kinds of arguments are made to securitize industries, to contribute to exceptionalizing. Typically when an issue is securitized, it’s like, okay, this has an elevated status, and it’s an issue now of national security, and so it’s above the law again.
China’s industry is more regulated than ours. Yes, we all know that trade competition and security interests are a thing. But this idea that the only way we can beat China is to let AI companies govern themselves and just make as much money as possible is not going to keep anyone safer.
Caroline: You’re in a David and Goliath position. You’re new, and your opposition is led by venture capitalists, as you say, and tech leaders. So I imagine you’re being vastly outspent. What do you feel like money can’t buy in this situation, in terms of what you’ve seen?
Ellen: Weirdly, one thing is LTF hasn’t spent the way that we expected them to.
[Money can’t buy] credibility, trust. I think you could see that was crystal clear with the data center fights. Counties regardless of political leaning, rural, urban, where people are rejecting data centers are being basically asked to name their price and saying no. And I think that’s because there’s no value proposition being provided.
Caroline: So some organizations, they might come to AlignIQ because they think, “We can’t afford not to adopt.” And of course our first response is you can’t afford not to adopt a policy. But what would make a consultancy like mine less part of the problem and more part of the fix?
Ellen: You need a certain amount of media literacy, tech literacy, to be able to even evaluate the work that these systems are doing. I feel comfortable trial-and-erroring. I don’t know which skills of mine I might be taking for granted when I’m able to do that, when I’m saying “Oh, this is easy” or not. But I think there is value in seeing, experimentally almost, what kinds of things might work for an organization. If I was trying to help an organization adopt AI, I would look at what their privacy needs are, what kind of information can and cannot be fed into a model, and what are the potential consequences of the model getting something wrong.
You’ve probably seen the studies that AI has not saved work, it’s intensified it because it is so output-driven. I think there’s alignment work to be done for teams: what do we want this to accomplish in our workplace, and what do we not? Sometimes when people adopt AI it’s like, we want to triple our productivity. I don’t think that’s right. For me it would be to improve the quality and experience people have at work. AI is a tool and workers are still the ultimate owners and accountable for their work. AI is not replacing thinking.
Caroline: We are B2B, and you work in a vaster system of different expectations. But what’s one AI guardrail that you would put in place tomorrow?
Ellen: I, personally, would say probably all systems need to be independently audited, as a baseline. Open and audited. No self-grading on homework, as they say. That would be a first step.
Wondering why so many people don’t trust the people building AI? “Know, Like, and Distrust: Why AI is so Unlikable”
Curious how OpenAI’s founders ended up in court? “Elon Musk v. OpenAI”
Know who else sat down with us to talk about AI safety? “Train Your AI First: A Conversation with Dr. Lu”
[In this post, we used AI for polish, not purpose.]
Leave a Reply