Nikhil Chandhok, chief product and technology officer at the stablecoin company Circle, said his company has not yet worked out what happens if AI agents get legal personhood. What it has thought about, he said, is agents acting on behalf of people. He then sketched a middle ground: a company that is "AI incorporated, AI managed, but had a board of directors that was human." The next step, he said, might be a board made up of other AI systems, and that is "not too hard to imagine."
The exchange took place on stage at Moonshots Live 2026, recorded on September 25, 2026 and released as an episode of Moonshots with Peter Diamandis on September 29. Host Peter Diamandis, founder of XPRIZE and Singularity University, shared the stage with Emad Mostaque of Intelligent Internet, ARK Invest's Cathie Wood and Chandhok.
The prompt: Argentina and "regulatory arbitrage"
Diamandis put the question to Chandhok by recalling an announcement by Argentine President Javier Milei, which the podcast had discussed about two months earlier. As Diamandis described it, Milei said Argentina would change its laws to attract AI companies and give agents personhood. Diamandis expected countries to compete this way, offering a kind of "regulatory arbitrage": rules loose or novel enough to pull companies across borders.
Milei's own statement is narrower than Diamandis's summary. In an official communiqué from June 18, 2026, Milei argues for giving legal personality to limited-liability companies that are run autonomously. He does not propose personhood for individual agents. His argument rests on identifiable assets and enforceable liability. Someone with a claim could go after the company's property, and authorities could sanction or dissolve it. Incorporation, he argues, would bring these firms under ordinary commercial law. This is close to the arrangement Chandhok went on to describe.
A company run by AI, overseen by people
Chandhok compared his middle ground to an ordinary corporation. AI would form and run the company, and humans would sit on the board. He listed the questions such a company raises: how it gets incorporated, what economic output it produces, "who holds the liability for the mistakes it makes," and how it distributes its profit. In his current thinking, he said, ownership and board seats in such a company stay with humans.
He called the arrangement "very imminent" and thought the money side could change soon. He said it was not hard to picture an AI using crypto networks to raise money directly, finding shareholders willing to fund it so it can "go and act in the world." His timing: "Probably this decade." He found it harder to guess what such companies would work on: maybe medical problems, maybe something else.
What DAOs left behind
The conversation then turned to decentralized autonomous organizations, or DAOs. These are online groups that use blockchain-based rules and token votes to manage shared money. The argument was that DAOs had worked out how to govern a group of people who did not know each other and lived in different countries. Agents are similar: they operate globally and are strangers to one another. DAOs flourished about four or five years ago before going out of fashion, but the discussion held that their record of what worked and what failed is prior art that agents could learn from.
One experiment from that period was quadratic funding. In Ethereum's explanation, people donate to projects during a funding round, and a matching pool is then split according to how many separate people backed each project, not just how much money came in. A project with a hundred one-dollar donations can end up with more matching money than one with a single $10,000 gift. The point raised on stage was that crypto already has tools like this that let parties who do not trust each other coordinate.
Personhood came back into the discussion as a way to make agents traceable. If an agent had legal status, others could check where it came from, who granted that status and why. The discussion also anticipated funding models native to the internet, not a trip to the bank, including the possibility that an agent might issue its own token.
From DAOs to "decentralized intelligent organizations"
The discussion also named a weakness in DAOs: they lacked intelligence beyond that of the humans running them. Adding AI could turn them into "decentralized intelligent organizations." That led to a question for Chandhok, who had moved to Circle from Meta: wasn't a digital economy of humans and AI entities what the metaverse was supposed to be?
Chandhok said he worked on Meta's augmented-reality glasses. He called the Ray-Ban glasses a good first step and the Orion prototype a good proof point. Meta unveiled Orion in 2024 as glasses that place digital information and contextual AI over the physical world. For him, though, the metaverse was never mainly about 3D graphics. He pointed to Pokémon Go, where players walk the real world looking for creatures they cannot see in advance, as an example of meaning layered on physical places. He said AI entities already offer meaning and connection, and he expects "new, weird societies" to emerge that are hybrids of humans and AIs.