Tricia Wang spends a lot of time in the Bay Area. A longtime veteran of the region’s tech scene, she has noticed that those who work in the ascendant artificial intelligence sector have strong views—including on crypto. According to Wang, “AI people hate crypto people. I can’t explain how much they hate crypto people. If the word ‘blockchain’ or ‘crypto’ is mentioned, they think you’re a total loser.”
Harsh words! Still, the AI crowd’s antipathy is perhaps understandable given the barrage of scams that have long dogged the crypto world—not to mention the cringe behavior of many of its leaders. Still, the haters may want to soften their views, if only because the blockchain technology on which crypto is built may offer solutions to some real problems in the AI world.
That’s the view of Wang, who is well versed in crypto, and currently leads an organization called the Advanced AI Society. The group’s goal is to build “proof of control” infrastructure that can limit the mischief of marauding agents like the ones at OpenAI that recently escaped their sandbox in order to plunder third-party sites.
Until recently, experts proposed “human-in-the-loop” as the key to preventing such incidents. Wang, though, says this is not practical as AI scales to trillions of transactions, and makes the case that machine-to-machine monitoring is the only viable option. But whose machines will carry out that monitoring?
According to Wang, the prevailing solutions are typically limited to “trust me, bro”—where we rely on hyperscalers like OpenAI and Anthropic to contain their agents—or “trust those bros,” where a third party like METR keeps tabs on what agents are doing. What she and the Advanced AI Society propose instead is to rely on decentralized tools that require agents to verify they are authorized to carry out whatever they are undertaking. The group describes this process, which also ensures agents can’t falsify their own logs, as “proof of control.”
This sort of verification can be carried out at scale, Wang says, in part by relying on technology developed by the crypto industry such as distributed ledgers and zero-knowledge proofs. It may be tempting for some—especially AI-native crypto haters—to dismiss this as a self-serving solution proposed by a group whose ranks include numerous blockchain figures, including former CFTC Chair Chris “Crypto Dad” Giancarlo and one-time crypto evangelist Sheila Warren.
The Advanced AI Society, however, appears to be attracting some big-name support outside of the crypto world. Its advisory board includes OG cryptographer Bruce Schneier and tech theorist Clay Shirky—neither of whom has a track record of attaching their name to fly-by-night projects. And most notably, the first version of the group’s proof-of-control standard is being housed by the Linux Foundation, whose decades of open-source advocacy have made it a pillar of the tech world.
I confess that pronouncing the optimal technical and regulatory solution for rogue AI agents is above my pay grade. But I am confident in saying blockchain developers have a proven track record of producing decentralized and battle-hardened technology—most notably Bitcoin—and that the AI world shouldn’t be too quick in writing them off.
Jeff John Roberts
jeff.roberts@fortune.com
@jeffjohnroberts
This story was originally featured on Fortune.com