The AI boom of the past year and a half has demonstrated the importance—and difficulty—of regulation in the tech sector. Striving for a regulatory environment that fosters innovation while addressing potential risks demands collaboration between governments and industry stakeholders. But such an equilibrium is exceedingly rare in the warp-speed world of Silicon Valley.

At the Fast Company Grill at SXSW, Frances Haugen, founder of Beyond the Screen; Evi Fuelle, global policy director at Credo AI; and Philip Martin, chief security officer at Coinbase took to the stage to discuss a path forward.

Thinking “locally”

Martin sees regulation as “an outgrowth of culture.” That is to say, a more realistic approach to regulating technology would be to think on a smaller scale.

“Is there a global regulatory standard? Probably not for any technology,” Martin said. “All politics is local, so we’re going to see that reflected in different countries’ or states’ or regions’ approach to regulation. If we start there, then what we need to go back to what are the first principles behind this and how can that be applied maybe in different ways in different jurisdictions.”

“An example would be as FTX blew up, different regulators needed the ability to protect their constituents,” he continued. “And that needs to be done in a way that also respects the need of a global business.”

Balancing the asymmetry

One thing Haugen is thinking about is the disparity between regulating the tangible versus intangible. For example, if you buy a car you’e able to test if the brakes work. Or if you buy produce, you can determine up front if it’s
fresh or not.

“When it comes to these technologies, we have to trust the companies that they are complying with a level of safety that we want. Or that when they’re facing trade-offs between safety and public safety, that they’re being resolved in the interests of the public,” Haugen said. “Given that we are moving from a tangible economy to an intangible economy, how do we balance that information asymmetry in a way that we get to ask questions or formulate questions ourselves and not just rely on these companies? And that means [tech companies] proactively saying, here’s how we’re approaching safety.”

Treating regulation like innovation

Fuelle is fully onboard with regulating AI—she’d just want that regulation to be as dynamic as the technology itself.

“Really what we should encourage is policy making that can be more of a testing process like we see in innovative industry. So allowing for policy that isn’t perfect to happen and thinking about what are the things that we can require of enterprises right now?” Fuelle said. “Can we require mandatory transparency reporting? Can we stipulate what data needs to be provided in order to demonstrate how the model was trained? There’s so much more we could be doing. Transparency sounds like a great concept, but most enterprises are pushing back on that. And I think that we need to reiterate that it’s very possible to introduce transparency to AI models. There’s algorithmic impact assessments, model cards, dataset cards. There are many ways that we can increase the understanding in this space.”