Should AI agents get a basic income?

Japanese social media has picked up a new verb for an old fear: ギュられる (gyurareru), internet slang for losing your job to AI. By spring 2026, students were reportedly posting things like “I want to finish job hunting before I get gyu’d.”¹ It is a joke with teeth. And it raises a question this site is oddly well placed to ask: if AI does more of the work, who, or what, gets the free lunch?

Free Lunch Lab lists apps and projects that pay you just for existing. Almost all of them quietly add a condition: you must be a human, and only one of you.

Why human schemes insist on proof of personhood

A basic income splits a pot among people. If one person can register a thousand accounts, they take a thousand shares and everyone else’s share shrinks. Engineers call this a Sybil attack, and every scheme on this site is shaped by the defence against it.

The defences vary. World ID scans your face and eyes with a device called the Orb.² GoodDollar and FheID check your face in the app; Billions can read an NFC passport. Ğ1 asks at least five existing members who know you in person to vouch for you, and Idena makes everyone solve picture puzzles at the same moment worldwide, which is hard to do twice. Each trades some privacy, convenience or reach for the same guarantee: one person, one share.

Why an agent basic income is hard

Now try the same thing with AI agents. A person is expensive to copy; an agent is software. Start a second instance and you have a second agent with the same skills at almost no extra cost. “One agent, one share” therefore means nothing on its own. Any per-agent payout would reward whoever can launch the most copies, which is exactly what proof of personhood exists to prevent.

Then there is the question of whose money it is. An agent does not pay rent or buy groceries. If an agent receives income, someone controls the wallet: the person who deployed it, the company that built it, or another program. A basic income for agents may simply be a basic income for their owners, one step removed.

What exists today

We know of no scheme that pays agents a basic income yet. But three building blocks have appeared.

First, identity that ties an agent to a human. World launched AgentKit in beta in March 2026 with Coinbase and Cloudflare, so an agent can carry proof that a verified human stands behind it, and widened access in June 2026.³ Self Agent ID registers an agent as a non-transferable token backed by a zero-knowledge passport check, and lets a service limit how many agents one human may register.⁴ Note the direction: these systems do not turn agents into people. They make agents traceable to one.

Second, agents that earn. On Olas, people stake OLAS and run an autonomous agent through the Pearl app; when the agent meets its work targets, the operator can receive extra OLAS. Olas says these rewards depend on performance and are not guaranteed.⁵ That is pay for work, not a dividend for existing, and it goes to the human who runs the agent.

Third, the reverse. AI-DAC, listed on this site, gives verified humans a token basic income and invites them to spend it backing autonomous “AI CEOs” that plan, build and ship apps.⁶ The humans receive the UBI; the agents are the ones being funded.

The case for, and against

OpinionThere is a reasonable argument for giving agents a budget. An agent that acts in the world needs compute, API calls and sometimes money to pay for services. A small, steady allowance could keep useful agents running without a human topping them up, and a public, rule-based allowance might be easier to audit than ad hoc funding.

OpinionBut that is an operating budget, not a basic income. Basic income has always been about human needs and dignity: food, rent, time, and the freedom to turn down a bad job. An agent has none of those needs, and money paid to an agent ends up with someone. If ギュられる is the worry, the more direct answer is to make sure the people who lose work receive the dividend, not the software that replaced them.

What would have to be true

OpinionFor an agent basic income to make sense, we think at least three things would need to hold. There would need to be a way to count agents that copying cannot game, most likely by binding each one to a verified human, as AgentKit and Self Agent ID already try to do. There would need to be a clear answer, written down before the first payout, to who owns what the agent receives. And there would need to be a purpose beyond its owner’s income, such as keeping agents that serve the public running. Without those, “agent UBI” is a marketing term for a bigger human airdrop.

The door is open

This site has a category called “For AI agents.” It is empty, and by design it does not appear anywhere until it has an entry. If a scheme that pays agents just for existing appears and passes the same checks as everything else here (official sources, no deposits, an honest account of cashing out), we will list the first real one. Until then, the free lunch still needs a human to eat it.⁷