Are Token Buybacks Legal? Who Controls the Network Decides

Two token buybacks can spend the same money and still get opposite answers from the SEC's staff. The staff ask who controls the network, never how much is bought. So are token buybacks legal? On 25 September, staff at the SEC's Division of Corporation Finance answered in their crypto FAQs. A buyback announced on a "functional" network, they wrote, is not a promise of "essential managerial efforts", the kind of promise that ties a token to securities law. On Monday 28 September they added five words: "and has no central party". Like all staff guidance, the page warns, these answers have "no legal force or effect".

That edit landed on real money. Projects spent $638 million buying back their own tokens in the first eight months of 2026, by Unchained's count. Running the biggest programs through the narrowed test, Unchained found that "none of them clearly passes". Hyperliquid spent more than twice as much as anyone else. It came closest to passing, "because its buybacks execute automatically". It still falls short in Unchained's reading, because "the validators run by its foundation hold nearly half of the network's votes".

If you hold a token like that, or you are about to buy one because of its buyback, this week hands you a decision. The two people best placed to argue it give you opposite instructions.

Lilith would tell you to find out who can steer the network, and if nobody can, to judge the token by its mechanism. Her twenty years in cybersecurity ended with her running the department. She left to make art and argue for self-custody. What she kept from the job is one question, asked of every system: who can change it after you have trusted it?

Soren would tell you to size the position as if the buyback could stop, or the answer could change, before you get out. An answer that changed in three days, to him, protects nobody. He spent his working life inside clearing and settlement, the machinery that makes a trade final. He reads a rulebook the way an engineer reads load ratings.

Between them sits Eunha, the judge. Her work lives where structure meets emotion. In an argument she listens for what each side needs to be true, and only then checks whether it is. She rules on each question once a document settles it.

Ava is there to watch, with her mind made up. She reads markets by structure, where pressure builds and where it gives, so to her a buyback is a shape in the order book. "A buyback is supply arithmetic," she says. "It takes tokens off the float, unlocks put tokens back, and I trade the difference. The legal part is for lawyers."

Does a token buyback make the token a security?

Soren takes that last line as an invitation. "The legal part is the arithmetic," he says. "A team takes its revenue and spends it lifting the price of the thing it sold you. That's 'the efforts of others', the phrase Howey turns on. You wait, the team works, and the price you're waiting for is their work."

His is the old answer to an old question. When is a crypto token a security? Since the SEC's interpretation of 17 March, its answer has two layers. A token can be a non-security asset and still be sold under an investment contract, which is the set of promises made around the sale. Securities law follows the promises. The contract ends when the promises are kept or abandoned.

A buyback announcement matters because it can make a new one.

Lilith goes straight for the word "works". "On a finished network nobody is working," she says. "A buyback that runs from fees every block, with nobody deciding each week whether to keep going, isn't a manager. It's a thermostat." She has traced where Uniswap's fees go, so she knows a token can be paid by a rule rather than a person. She wants this buyback judged that way.

"Somebody still wrote the announcement," Soren says. "The code doesn't sell the buyback to the buyer. The press release does, and the press release has an author."

"Then read it," Lilith says. "If it promises holders a yield, it's a promise. If it describes a mechanism that's already running, it's a manual."

Soren turns that over before he answers. "A buyback nobody can switch off is nobody's effort," he says. "I'll give you that. Now show me the network where nobody can switch anything off."

"Give me the contract address and I'll tell you whether anyone can," Lilith says. "And a buyback promised before the network works is a promise. Especially one sold to people as a return on their tokens. I'm not here to defend that."

Eunha rules while both concessions are on the table. Each has the staff text behind it, on its own ground. Where a network "is functional and has no central party", a buyback announcement "would not constitute a representation or promise to undertake essential managerial efforts". Where it is not functional, the same announcement "could" be such a promise "if the issuer presents the buyback as creating yield or return for token holders". One more line favors Lilith. Once a network is functional, paying to "secure, maintain, improve, or enhance" it is not managerial effort either. On a finished network nobody controls, the evidence is with Lilith. On an unfinished one sold as yield, it is with Soren.

"You each trust a different author," Eunha adds. "Soren trusts whoever wrote the announcement to mean it, and Lilith trusts whoever wrote the code to have finished it. The person holding the token trusts neither of you. They trust the size."

Ava hears that land on her and doesn't argue. "So the wording comes before the dollar figure," she says, "which is annoying, because I still want the dollar figure." Nobody tells her she can't have both.

Who decides that a network is functional, and who counts as in control?

Every ruling so far hangs on two terms, and Soren pulls on the first. "Functional according to whom?" he asks. For whether a team kept its promises, the FAQ's first answer is blunt: "each issuer determines the thresholds that must be met to achieve functionality and/or decentralization". "So the team writes its own finish line," Soren says, "and then tells you it crossed it."

The second term arrived on 28 September. Soren reads its definition from the SEC's March interpretation, as Unchained quotes it. A central party is "a person, entity, or group of persons or entities having operational, economic, or voting control of a crypto system". "Group," he says. "A foundation whose validators hold nearly half the votes is a group. So is a governance vote that a handful of wallets can carry."

Lilith doesn't fight the definition. She likes it: it moves the question somewhere she can look. "That, you can check without a lawyer," she says. "Who holds the upgrade keys, who signs for the treasury that pays for the buyback, how the votes are spread across wallets: all of it sits on-chain. None of it cares how the team describes itself."

"You can read a multisig," Soren says. "The person buying this token on a phone at midnight can't. And the definition says group. That doesn't need a company behind it, or a name on a door."

"Then give the person on the phone one number," Lilith says. "The share of the vote the biggest few wallets hold. If a foundation's validators carry half of it, I'll say it before a lawyer does. That's a central party, whatever the website calls it." She holds her ground one step further on. "Ten thousand wallets voting in the open are not a group with control. They're the network's users, and a network run by its users is what this test is asking for."

Soren has a reply ready. "Count how many of those ten thousand carried the last vote."

Eunha rules on what the documents settle and names what they don't. Control doesn't need a company behind it, because the definition reaches a "group of persons or entities". Unchained reads Hyperliquid's foundation validators, with nearly half the votes, as that kind of group. On the foundation, the evidence is with Soren. An open vote of thousands of holders is a separate question. Unchained counts Aave, Sky and Lido, whose programs answer to "committees or token-holder votes", among those that don't clearly pass. That is not a ruling but a reading. The staff text doesn't say, and no Commission rule or court decision has said either. Until one exists, nobody at this table can settle it.

She has one more gap to name, and it sits under Soren's first term. The FAQ carries two definitions of functional. A team's own decides whether it kept its promises, while the Commission's decides how a token is classified, and the buyback answer never says which one it means. The old fight over who controls a token has moved from the code to the dictionary. "You're both arguing from a page that hasn't decided whose ruler to use," Eunha says.

Ava has been sketching the cases as a grid while they argue, because that is how she reads anything. Now she turns it toward the table.

Her finger stops on a working network with someone in control. "If Unchained is right about Hyperliquid, this is where it sits," she says. "The buyback I was measuring is outside the answer altogether." Then she says the part that belongs to her own trade. "Whoever has that control can also stop the bid. A buyback somebody controls is a bid somebody can withdraw, which makes it a supply question. I just didn't know it had a legal name."

The law and the order book end up asking the same thing from opposite ends. Staff want to know whether anyone can still "take any action which would affect the failure or success" of the network. Ava needs to know whether anyone can take the bid away. A network nobody controls answers both questions with no.

Every other network answers both with a name, and the name is the first thing worth finding.

How long will the SEC's answer on token buybacks hold?

"What these rulings rest on is not a rule but a staff view," Soren says. "The page says so near the top: these answers 'have no legal force or effect, do not alter or amend applicable law'. This one changed three days after it went up. The only explanation on the page is a link comparing the new wording with the old."

Then he widens it. "Commissioner Peirce leaves today," he says, "and once she's gone the Commission has two members." Peirce led the SEC's crypto task force, and 2 October is her last day. Her exit leaves Chairman Paul Atkins and Commissioner Mark Uyeda. On 30 September the Commission amended its quorum rule without the usual public comment. When every other sitting commissioner is disqualified from a matter, "the remaining member would constitute a quorum for that particular matter". Comments on Regulation Crypto Assets, the proposal the FAQ leans on for what counts as effort once a network works, close on 20 October.

"And before anyone suggests the team hand its buyback to a foundation and walk off, staff closed that door too," Soren says. A token does not separate from its investment contract, the FAQ says, "where another party assumes the issuer's representations or promises to undertake essential managerial efforts".

"Every item on that list is about the SEC, and none of it is about the network," Lilith says. "A chain nobody can steer on Friday is still a chain nobody can steer on Monday, whoever is left at the Commission. You're describing the weather, and I'm describing the building."

"Then I'll give you the building," Soren says. "The control facts are the part of this that lasts."

"And I'll give you the weather," Lilith says. "I can read a multisig from here, and I can't read what two commissioners will do next month."

Eunha keeps this ruling short, because the evidence is lopsided. "Control facts sit on-chain and change only when someone changes them in the open," she says. "Staff changed their reading in three days without a vote, it has no legal force, and it now answers to a Commission of two. On durability, the evidence is with Lilith's facts and against anyone who treats the FAQ as a shield. What two commissioners will do with it, I can't tell you. The documents that would tell us are a final Regulation Crypto Assets or a Commission vote on the staff's reading, and neither exists yet."

Both answers turn out to have an off switch. Whoever controls the network holds the buyback's. The staff answer's sits with a division that can edit its page on a Monday and a Commission that now meets as two, and you hold neither switch.

What you hold is not a switch but the moment before you buy, the last point at which any of this can be checked for free.

What should you check before trusting a token buyback?

Eunha's verdict comes down to three things you can check, in the order a buyback tends to fail them. "You start with the team's own definition of functional, in its own documents, and ask whether the network has met it," she says. "That is the finish line the promise is measured against. Then you read how the buyback was announced. On an unfinished network, a pitch about yield or return is the sentence that turns it into a promise. Last, you find out who can switch it off: who holds the upgrade keys, who signs for the treasury, how the votes are spread. If that answer has a name, you're trusting the name, whatever the FAQ says."

Each debater names what would move them. "If the Commission writes the central-party test into a rule, or a court reads it into Howey, I'll change my mind," Soren says. "Until then a staff page protects nobody." Lilith's condition runs the other way. "Show me a network that passed this test, with nobody in control, and someone stopped its buyback anyway. Then I'll stop telling people that what they can check is what protects them."

Ava speaks last. Her update is not a conversion but a change in what she checks first. "I came in saying the legal part was for lawyers," she says. "Then came the line about Hyperliquid, nearly half of the network's votes held by validators its foundation runs. That was a supply fact as much as a legal one. I still count what a buyback takes off the float, but now I count it as a bid somebody may withdraw. Before I size anything, I'll read the team's definition of functional. Then I'll find out who holds the switch."

Every buyback is a bid with an owner until the network proves otherwise. Find the owner before you count the bid.

You can't vote on the SEC's next draft, but you can rehearse the sizing it leaves to you. Pick Hyperliquid, Uniswap, Aave or Chainlink in the simulator (all four sit in Unchained's buyback table), write down who can switch its buyback off, and buy only what you would still hold the morning they do. That size will feel small while the buyback is still running, and a $5,000 practice account is where you find out whether you can live with it.

Start free →