What AI regulation and ownership could be

As an ordinary internet user, I am skeptical about AI. My colleagues and I have reported on several harmful aspects of the technology in the US: threats to democracy, its impact on the environment, mental health, the workforce, military intervention—the list goes on. 

While researching the gaps in conversations around enforceable government regulation of AI startups, both big and small, eager to innovate and launch new products, I came across AI researchers, policy experts, and even “ordinary internet users” like me trying to reimagine how the industry could work. If AI is unstoppable, can we build it in a way where the public has more of a say in how the technology is made, used, and who it benefits? 

To answer these questions, I spoke with B Cavello, the director of emerging technologies at the Aspen Institute. Cavello wants to make sure that ordinary users like you and me have some say in how this powerful technology is developed and regulated. I talked to Cavello about the gap between the recent proposals among tech executives and lawmakers to give Americans financial ownership stakes in major AI companies and what actually ensuring public safety could look like.

This interview has been condensed and edited for clarity. 

I’ve heard a lot of coverage on proposals of public ownership of AI from folks like OpenAI CEO Sam Altman and Sen. Bernie Sanders. Are there other prominent proposals being discussed?

Ownership is complicated. I’m a big proponent of the movement for public AI. Public AI is open source, centering public access, accountability, and sustainable public goods. That means making sure that not only do people have free access to things, but also access to build models in their own local context. 

When we say accountability, we don’t just mean regulation on harms we’re trying to prevent. We also mean the capabilities that we want to get built in the world. When we talk about AI leading to a new age of human flourishing, we often aren’t talking about the basics, and that’s an important place to start when we think about public accountability and then sustainable public goods. 

When we build products within this framing, we want them to be essentially treated like public infrastructure. A lot of work in the public AI community has been on open source because that’s one of the spaces where people are able to continue to build on top of each other. CurrentAI [a public-private partnership of governments, companies, and organizations to fund public interest tech] announced in the last couple of weeks their AIPotluck initiative, which is bringing together different open-source AI builders to build a full open stack [meaning making the entire AI development process accessible to the public].

It’s less about trying to take the market share or revenues of big AI [companies, similar to Altman’s and Sanders’ proposals], but more that we’re going to distribute these capabilities, such that now they aren’t the only players in town—and we’re distributing [electrical] power that way. 

Who are the types of people and communities that are participating in public AI discussions?

There’s a big movement around AI sovereignty across the African continent and Europe about trying to disentangle from American AI providers. To that end, ownership is about local representation and empowering a local tech community. There’s a conversation in the US, especially around a potential AI bubble, whether or not it might burst, and how we might have public ownership or public uptake of these technologies. 

As you alluded to, there are folks like Sen. Sanders who are responding to the concerns around AI safety and power concentration, who are trying to look for interesting and creative avenues for trying to have some sort of power over these companies through partial ownership of them, but I think is like ultimately not where I would focus, especially per some of the conversations around the uncertainty of the [economic] future of these systems. It is sort of accepting defeat on other forms of regulation to say we’re just going to be passengers on this ride. I think Sanders’ team and a lot of people working in Congress don’t want that to be the only strategy going forward.

Windfall Trust, [a network of independent AI researchers and strategists], has been working on how to encourage people before a crisis happens [like mass unemployment] while the pressure is low, and get some pre-commitments on trigger conditions for different [ameliorating] policies. 

Considering all the discussions around reimagining AI, how would these ideas become policy?

The thing I parallel it to is the renewable energy transition, where on one hand, we needed the tech to be there. We needed solar panels to get super good and cheap. But we also needed rebates on the policy side that enable us to leverage those technologies. 

Congress and, broadly, governments tend to be reactive institutions. 

The public is also becoming more aware and engaged on issues around AI. So some of these things will be showcased at a more local level before they make it into larger policy.

As a regular internet user, what are things that I should be thinking or concerned about?

How do we establish public accountability in how data centers are operated? For the last 20 years, there have been people demanding transparency in technology companies and basically being told we can’t do that because of the First Amendment. We have adopted this mentality that tech companies can be secretive—we can’t ask them anything, and they don’t divulge anything. We can absolutely require reporting about some of the ways data centers are operating.

I would love to see people pushing on public access to data centers. Let’s demand that we have set-asides that 20 percent of a data center is allocated toward publicly accessible, accountable, and sustainable public goods. 

We already have the tools. We’ve seen a vast hollowing out of the apparatus that is meant to hold these entities accountable. Federal agencies are meant to do this work—the IRS. If somebody in the public cares about holding big AI companies to account, part of [the public effort] means championing the people who can do that work and encouraging the election of people who are interested in that. 

I would encourage people in their own use of AI technologies to think about how to use open systems and run things locally. There are some folks who are engaging with public libraries across the country, where people can access these tools for trying things out. OpenAI and Anthropic don’t monopolize what is possible [with AI].

Is there something analogous to conversations that happened previously on how to think about what AI could be? 

I would like our AI future to look more like electricity: a reliable, publicly accountable utility that has universal service requirements to make sure that everyone can be connected. And at the same time, we have batteries and home generators [as individuals]. 

It’s not to say that there can’t be interesting innovations on top of existing infrastructure. In many ways, I think of AI as the cool thing we can do with electricity. Here we are having a Zoom call—what an innovation in electricity! We should have public utilities, and we should have a thriving market of all different kinds of community solar and private companies that’ll build innovative batteries, but recognizing that we all deserve access, we all deserve accountability. We deserve, if we want, to have an innovation ecosystem. We need a sustainable infrastructure layer that people can count on and build on top of, and not be worried that [Anthropic] Mythos-style, suddenly everything is revoked. 

There are things that are not great about our public utilities. People are having fights about data centers in part because of the way our electrical grid is run [such as prioritizing data centers over local residents]. So I don’t think that it is a perfect one-to-one.