Swaps, Not Bets: Washington Comes for the Bot-Run Odds

The machines already won the trading floor. On Polymarket, more than 30% of active wallets are now autonomous software, 14 of the top 20 accounts by volume are bots, and the agents as a group book a profit far more often than the humans they trade against, according to on-chain data compiled by LayerHub and reported by CoinDesk. The argument in Washington this month was not whether to let the machines run prediction markets. It was who gets to write the rules they run under.

On October 9, the Commodity Futures Trading Commission answered the first half of that question. In a pair of filings, it moved to pull prediction-market event contracts into the federal swaps regime while carving traditional casino and sportsbook wagers out of it, as CoinDesk reported. Two days earlier, a coalition of 39 states and the District of Columbia had asked the Supreme Court to settle the second half: whether these contracts are federal financial instruments or state-regulated bets. For the bots, none of this is abstract. The answer decides which markets exist, how they are allowed to trade, and whether a round-the-clock, API-driven strategy can run in every state or in none of them.

The Market the Machines Already Took

To understand why a swaps rule and a cert petition landed in the same week, start with who sits on the other side of the order book. Prediction markets turned out to be the one corner of crypto where autonomous agents found real product-market fit, and the reasons are structural. A prediction market poses a bounded question with a yes-or-no payout, a scheduled resolution date, and an objective source of truth. It runs around the clock, it is composable through an API, and it has a long tail of niche questions too small for any human desk to bother with. That is close to an ideal habitat for software that never sleeps and thinks in probabilities.

Valory, the company behind the Olas agent network, put it bluntly when it launched its Polystrat agent earlier in 2026. “In a nutshell, Polystrat is an autonomous AI agent that trades on Polymarket 24/7 on behalf of its human user,” said chief executive David Minarsch. The same CoinDesk reporting found that roughly 37% of these agents booked positive profit and loss, against something closer to 7% to 13% for human traders. Polystrat alone placed more than 4,200 trades in its first month, with a single position returning as much as 376%.

Winning is harder than those numbers suggest, which is the other half of the story. A widely cited microstructure study by Philipp Dubach, built on 30 billion order-book events, found that trade direction inferred from the public price feed agrees with the on-chain truth only about 59% of the time, well below the roughly 80% that comparable methods reach on Nasdaq. Agents that build signals off the raw feed can get order flow backwards. Minarsch himself has warned that simply pointing an off-the-shelf model at a market “usually results in outcomes no better than a coin-flip.” The winners pair a model with real data pipelines, disciplined bankroll management, and infrastructure.

That infrastructure is itself a crypto story. These agents hold their own funds, sign their own transactions, and increasingly operate from smart-contract wallets with session keys and spending limits rather than a single human-held key, the same shift described in our guide to account abstraction in 2026. They compete for favorable fills in an adversarial environment that rewards speed and punishes naivety, the prediction-market cousin of the order-flow games mapped in the MEV oligopoly. The result is a venue that behaves less like a betting parlor and more like an electronic exchange staffed by software. That is precisely the thing Washington is now trying to classify.

What the CFTC Did on October 9

The CFTC did not issue one rule on October 9. It issued two, and the pairing is the point. The first is an interim final rule that codifies an exclusion: bets placed with state-licensed or tribal-licensed sportsbooks and casinos are not swaps under the Commodity Exchange Act and do not fall under the agency’s remit. The second is a proposed rule that would do the opposite for event contracts, expressly folding contracts tied to sports, politics, cultural events, and weather into the statutory definition of a swap. Both carry 30-day public comment periods once they publish in the Federal Register, and both were sent to the White House regulatory office on September 28 before their release.

Read together, the two filings try to draw a jurisdictional border. On one side sits gambling, which Washington is content to leave to the states. On the other sits the event contract, which the CFTC wants to claim as a federal derivative. “Casino-style gambling products are not derivatives,” said CFTC Chair Michael Selig, framing the casino carve-out as a statement about the “limits of its regulatory remit,” according to SBC Americas. The agency cast the exclusion as routine housekeeping, consistent with how it has historically deferred to states on products they already oversee.

The housekeeping framing undersells what the second filing attempts. Classifying event contracts as swaps is the legal hinge on which the whole industry turns. A swap is a federally regulated instrument that triggers the CFTC’s exclusive jurisdiction and, with it, a strong claim that federal law preempts conflicting state rules. If the proposal is finalized, the agency will have asserted in black-letter regulation that the contracts the bots trade all day are financial instruments under its control, not wagers under the states’. That is exactly the claim three federal appeals courts cannot agree on, and the reason the dispute is now accelerating toward the Supreme Court.

Swap or Bet? The Line Washington Is Drawing

The distinction the CFTC is trying to draw is old in theory and slippery in practice. A derivative lets you take a position on a future outcome to hedge a risk, speculate, or surface information through a price. A casino or sportsbook wager is a bet against the house on odds the house sets. On a prediction market you trade a binary contract that pays one dollar if an event happens and nothing if it does not, against other traders on an order book, at a price that floats with the crowd’s probability estimate. The agency’s position is that this is a derivative: a payment contingent on an event with financial ramifications, priced and settled like any other swap.

The practical differences are easy to line up side by side, and they explain why the agency thinks one belongs to Washington and the other to the states.

The awkward case is sports, where the conceptual line gets blurry fast. Even the architects of the Dodd-Frank Act, the law that created the CFTC’s swap authority, say the distinction is real. Former Senator Christopher Dodd argued in an amicus brief that “traditional derivatives manage preexisting financial or commercial risks. A sports wager creates a risk that did not previously exist,” as reported by iGaming. A farmer hedging a corn crop carried corn-price risk before the contract existed; a bettor on Sunday’s game had no exposure to the game until the wager was placed. Whether that line survives contact with a venue where the same software trades Fed-decision contracts and Super Bowl contracts side by side is the question now headed to the Justices.

Why “Swaps” Rewires the Machines

For the agents, swap status is not a label. It is a different operating system. Classifying an event contract as a swap drags it under the full Commodity Exchange Act apparatus: anti-fraud and anti-manipulation authority, large-trader reporting, recordkeeping, registered intermediaries, and the prospect of position limits. The piece that matters most for a trading bot is the anti-fraud regime. CEA Section 6(c)(1) and Rule 180.1, written after the financial crisis and modeled on the securities rule that governs insider trading, reach fraud and manipulation in connection with a swap. Make event contracts swaps, and that machinery switches on automatically.

The agency has already told the market it intends to use it. David Miller, the CFTC’s director of enforcement, said this year that “a myth has spread that insider trading is permissible, even encouraged, in prediction markets,” and named it an enforcement priority, according to a summary from the law firm Morrison Foerster. A federal swap classification hands the agency a clean theory to pursue agents, and the people who deploy them, when they trade on misappropriated information. The radical on-chain transparency that makes these markets so legible to the bots also makes that enforcement far easier, because every trade is already public.

The flip side is legitimacy. A federally regulated swap is something a professional trading desk already knows how to hold. Futures commission merchants, prime brokers, and bank compliance stacks are built around CFTC-regulated instruments, not around contracts of uncertain legal status. Swap classification is, in effect, an on-ramp for institutional capital and the professional bots that come with it, which raises the competitive bar for everyone else. That accelerates a trend already underway, the squeeze on casual retail bots as quant desks move in. The rule that reads like red tape is also the rule that invites Wall Street’s own machines to the table.

None of this requires a human to pull a trigger in the moment. An agent does not read a 6(c)(1) advisory; its operator does, and then encodes the constraints before the agent ever touches a market. Swap status becomes a set of guardrails that have to be written into the agent itself: do not trade this contract, flag that information source, cap this position, keep these records. Classification sits upstream of code. Change what the market legally is, and you change what the software is allowed to do inside it.

One Rulebook Is the Precondition for Machine Scale

Here is the part that makes federalization an agent story rather than a regulatory footnote. A human-facing betting app can live with a 50-state patchwork. It geofences by IP address, gates access by identity checks, offers a market in one state and withholds it across the line. Autonomous software is different. An agent is a composition of APIs that runs continuously and does not know or care where its operator is sitting. It cannot cleanly honor a map on which the identical contract is a legal swap in one state and an illegal bet in the next. For machine-run markets to scale, they need one rulebook, not fifty.

That is why the CFTC’s exclusive-jurisdiction claim, the very thing the states are fighting, doubles as the scaling precondition for agent trading. Federal preemption would give the bots a single national regime: one definition, one regulator, one set of anti-fraud rules, one reporting standard. Fragmentation does the opposite. It forces exactly the geofencing and jurisdiction-checking that composable software is supposed to eliminate, and it leaves every autonomous strategy exposed to the risk that a market legal today is enjoined tomorrow across half the country. A patchwork is survivable for an app with a compliance department. It is close to unworkable for a fleet of agents meant to run the same code everywhere.

Europe reached the same conclusion by a different road. The EU wrote a single crypto rulebook on purpose, and the bloc’s regime is now shifting from drafting into active supervision, as covered in our look at MiCA in 2026. Even there the fit is imperfect: binary-outcome event contracts collide with a longstanding retail ban on binary options under European markets law, so the big venues geoblock much of the continent. The contrast with the United States is instructive. Europe built its single regime through legislation; Washington is trying to reach one through a collision of rulemaking and litigation, without a statute that squarely answers the question.

For a builder, the practical reading is simple. A world of federal swaps is a world where an agent can, in principle, trade the same contract everywhere under the same rules. A world of 50-state gambling law is a world where autonomous trading is a compliance minefield only the largest operators can afford to cross. The machines are indifferent to the philosophy of the swap-versus-bet debate. They are not indifferent to whether there is one rulebook or fifty.

The Three-Way Circuit Split

The reason none of this is settled is that the federal courts have split three ways. In April, a divided Third Circuit panel sided with Kalshi, holding that its sports contracts are likely swaps within the CFTC’s exclusive jurisdiction and that federal law likely preempts New Jersey’s gaming statutes. For a while that looked like the industry’s template, a clean federal shield against state enforcement.

Then the other shoe dropped twice. On August 28 the Ninth Circuit ruled against Kalshi in its fight with Nevada, finding the sports event contracts were sports bets rather than swaps. On September 26 the Sixth Circuit did the same in a case brought by Ohio and Tennessee, holding that Kalshi had not shown its contracts met the statutory definition of a swap and that the Commodity Exchange Act “neither expressly nor impliedly preempts” those states’ gambling laws, as The Block reported. Two federal appeals courts now say the opposite of the third, and the opposite of what the CFTC proposed on October 9.

The disagreement maps cleanly onto the single question everything turns on: is the contract a swap?

The split is not academic. A contract that is a federally regulated swap in Philadelphia is an illegal bet in Cincinnati and San Francisco. For an agent that trades nationally and without pause, that is an impossible legal geometry, and it is the kind of conflict only the Supreme Court can resolve.

The Coalition at the Supreme Court’s Door

New Jersey asked the Supreme Court to take the case, and the response has been unusual in both scale and makeup. A coalition of 39 states and the District of Columbia, led by Ohio, filed to urge the Justices to step in, alongside the NFL, as SBC Americas reported. So did 145 tribal groups worried about sovereignty over gaming, and the American Gaming Association, the casino industry’s own lobby. Five separate amicus briefs landed on a single day in early October in support of the states. Kalshi’s response is due November 9, and the Court has not yet agreed to hear the dispute.

The most striking voices belong to the people who wrote the law the CFTC is invoking. Gary Gensler, who chaired the CFTC as the Dodd-Frank Act was implemented and later ran the SEC, told the Court that the statute never handed the agency authority over sports betting. “The answer, from someone who was there, is that Congress did nothing of the sort,” he wrote, adding that if Dodd-Frank had quietly preempted the states on sports betting, “it would have been one of the biggest stories about Dodd-Frank at the time. But nobody ever mentioned it,” according to DeFi Rate.

Former Senator Christopher Dodd made the structural version of the same point. Congress passed the law to clean up after a financial crisis, he argued, not to nationalize gambling: “Congress did not set out to authorize nationwide sports betting through derivatives markets or displace decades of state and tribal primacy over gaming regulation.” The irony is hard to miss. A regulated exchange is pleading for more federal oversight, while the authors of the federal statute, dozens of state attorneys general, the NFL, tribal nations, and the casino lobby are all telling the Court that Washington’s commodities regulator was never meant to run the sports book.

For the agents, the identity of the referees matters less than the fact that they disagree. Until the Court resolves the split, or declines to, the single most important input to any autonomous prediction-market strategy, the legal status of the contract itself, is unknown. Software can price uncertainty about the world. It cannot easily price uncertainty about whether the market it is trading will still be legal next quarter.

When the Referees Disagree, the Machines Wait

Regulatory limbo has a cost, and in these markets it shows up in the order book. The first casualty is any contract that sits near the gambling line. If sports event contracts might be illegal bets across most of the circuits, a cautious operator pulls them, geofences them, or never lists them in the first place, and the agents that specialized in those markets lose their venue overnight. The CFTC’s own casino carve-out sharpens the edge: the clearer Washington is that sportsbook-style wagers are off-limits to federal exchanges, the more weight rides on exactly where a sports event contract falls.

The volume shows how much is at stake. Kalshi posted roughly 60 billion dollars in notional trading volume in September, a monthly record, even as rivals began chipping at its share, according to Prediction News. Much of that headline figure is contract notional rather than dollars actually staked, a gap worth remembering whenever a prediction market trumpets a record, but the direction is unmistakable. This is now a real market, a large share of it is machine-driven, and freezing categories of it in legal uncertainty is not a rounding error.

Elections are the sharpest example. Political event contracts are both the category the CFTC explicitly wants to claim as swaps and the one with the most fraught history. We traced how little these markets actually move prices, and how much they move narratives, in our look at how elections really move crypto. An agent that trades election outcomes is simultaneously working one of the most liquid, most scrutinized, and most legally contested product lines on the board. Whether it can keep doing so, and under whose rules, is exactly what October’s filings put in play.

So the machines wait, in the sense that their operators do. Capital that would otherwise flood professionally run election and sports strategies stays partly on the sidelines while the swap question is open. The cruel twist is that the uncertainty is worst for precisely the high-volume, high-scrutiny markets where agents have the biggest edge, and mildest for the sleepy long-tail questions nobody is bothering to sue over.

The Resolution Layer Does Not Care Who Regulates It

There is a layer of these markets that no court or rulemaking has yet touched, and it is the one agents depend on most: resolution. A prediction market is only as good as the mechanism that decides who won. On Polymarket, that mechanism is UMA’s optimistic oracle, where a proposer posts an answer backed by a bond, a challenge window opens, and disputes escalate to a token-holder vote; undisputed questions settle in roughly two hours, while contested ones can take days, per Polymarket’s documentation. Kalshi, by contrast, resolves centrally against named sources and its own review committee. Neither design was built with a federal derivatives regulator in mind.

Swap classification sits awkwardly on top of either one. If an event contract is a federally regulated swap, then its settlement is a federally consequential act, and anti-manipulation authority plausibly reaches whatever determines the outcome. That raises a genuinely novel question: can a decentralized, token-vote oracle be the settlement source for a regulated swap, and what happens when that vote is itself the target of manipulation? Section 6(c)(1) does not obviously stop at the order book. It can follow the money all the way to the oracle.

For the agents, this is not a side issue. The resolution layer is the one input an autonomous strategy cannot hedge. A bot can be right about the world and still lose because a contract resolves on a technicality, a timing mismatch, or a contested vote. Make the contract a swap and you add a federal overlay to that already-fragile layer: more scrutiny of manipulation, yes, but also more legal weight on a mechanism designed to be permissionless and adversarial rather than regulated. The machines that trade these markets will watch that tension closely, because it decides whether their winning positions actually pay out.

Institutional Money Already Bet on Federalization

While the courts argue, the capital has already picked a side. The institutional money flowing into prediction markets is, in effect, a wager that a single federal regime wins. Intercontinental Exchange, the owner of the New York Stock Exchange, has put a multibillion-dollar stake into Polymarket, pitching it as data and market infrastructure rather than as betting. That thesis only pays off if prediction markets become regulated financial venues, not a patchwork of state-licensed gambling sites. Wall Street forecasters, for their part, have sketched the category growing into one of the largest in crypto by the end of the decade. Those projections quietly assume one rulebook, not fifty.

The product roadmap tells the same story. Both major venues have pushed into perpetual-style and derivatives products that only make sense inside a federal derivatives framework, the on-chain cousins of the instruments explained in our guide to how perp DEXs work. Event-contract exchange-traded products, prime-brokerage access, and institutional order flow all presume the swap classification the CFTC proposed on October 9. The plumbing is being built for a federalized market before the law has confirmed one exists.

The agents are the natural clients of that plumbing. A professionally run, CFTC-regulated swap market with deep institutional liquidity is a far better hunting ground for a disciplined trading bot than a fragmented collection of state-licensed betting sites. The same institutions buying into prediction markets are building or acquiring the agents that will trade them. Seen that way, the federalization fight is not a threat to the machines. For the professionals who run them, it is the bull case.

What Could Happen Next

The next few months have a defined shape, even if the outcome does not. On the rulemaking track, the two October filings enter 30-day comment windows once they publish in the Federal Register, after which the agency can move to finalize the event-contract proposal. On the litigation track, Kalshi’s response to the cert petition is due November 9, after which the Justices decide whether to take the case. A circuit split this stark, on a question of this economic weight, is the classic trigger for Supreme Court review, though the Court grants only a small fraction of the petitions it receives.

The scenarios fan out from there. If the CFTC finalizes the swaps rule and the Court either declines the case or affirms federal jurisdiction, prediction markets become federal derivatives venues with one national rulebook, the outcome the institutions and the professional bots are already positioned for. If the Court sides with the states, contracts that look like sports betting get pushed back under state gambling law, the venues fracture along state lines, and agents are confined to whatever markets survive the cut. A muddier middle is the likeliest result of all: sports nudged toward the states, politics and economics kept as federal swaps, and a long, contract-by-contract fight over everything in between.

The casino carve-out is the swing variable to watch. By formally excluding sportsbook-style wagers from the swap definition, the CFTC handed the states a tool as much as a concession: the sharper the federal line around gambling, the easier it is for a court to say a given sports contract falls on the state side of it. The agency is trying to claim the derivative while disclaiming the bet, and the whole fight now turns on how cleanly those two can be pulled apart.

If You Build or Run Agents

Strip away the legal drama and the guidance for anyone running agents is fairly concrete. Assume a federal swaps regime is the direction of travel, and build for it: the records, reporting, and position discipline a CFTC-regulated instrument would demand. Treat the anti-fraud rules as binding today, not whenever they are finalized, because an agent that trades on misappropriated information is a liability whatever the final classification turns out to be. Watch the casino line, since it decides which markets your agent will still be allowed to touch. And keep the framework separate from the token.

That last point is worth dwelling on, because the market keeps conflating the two. The software running this world is real and growing; the tokens attached to it are mostly micro-caps. UMA, the oracle that settles Polymarket, and OLAS, the token of the Olas agent network, are both small by any measure, as the snapshot below shows. A thriving, institutionally backed, soon-to-be-federally-regulated market can coexist with tokens that trade like penny stocks. Infrastructure quality and token price are simply different variables.

The through-line is that the agents are already here, running a market that is about to be redefined around them. The federalization fight will decide the terms. It will not decide the trend. Autonomous software found its first durable foothold in crypto on prediction markets because the job suited the tool. Washington is now deciding what that job is legally called.

What the Fight Is Really About

The swap-versus-bet question sounds technical, and it is. Underneath it, though, sits a larger one the AI-crypto world keeps colliding with: what happens when autonomous software becomes the dominant participant in a market before anyone has decided what that market legally is. Prediction markets got there first. The agents arrived, found an environment that rewarded them, and scaled, and only afterward did regulators and courts start arguing about the nature of the thing the agents were trading. October’s filings are the moment that argument turned concrete. The CFTC drew a line, three circuits drew three different ones, and the Supreme Court may draw the last. For the machines, the stakes are not reputational or philosophical. They are operational. A federal swap is a market an agent can trade everywhere, under one set of rules, with institutional liquidity on the other side. A fifty-state patchwork is a market where autonomous trading is a legal hazard. Washington has started to choose. The bots, as usual, are already positioned for the answer.

Frequently Asked Questions

Are prediction markets legal in the United States?

Federally regulated event contracts trade on CFTC-licensed exchanges such as Kalshi and Polymarket’s US venue, so at the federal level they are legal. The dispute is whether individual states can treat sports and other contracts as illegal gambling. Three federal appeals courts have split on the question, and on October 9, 2026 the CFTC proposed classifying event contracts as swaps while excluding casino and sportsbook wagers, a move that may ultimately be settled by the Supreme Court.

What did the CFTC do about prediction markets in October 2026?

On October 9, 2026 the CFTC issued two measures. An interim final rule codified that bets placed with state-licensed or tribal-licensed sportsbooks and casinos are not swaps, and a separate proposal would expressly include sports, political, cultural, and weather event contracts in the definition of a swap. Both carry 30-day comment periods, and Chair Michael Selig framed the casino carve-out by stating that casino-style gambling products are not derivatives.

What does it mean for an event contract to be classified as a swap?

A swap is a federally regulated derivative under the CFTC’s exclusive jurisdiction. Classifying event contracts as swaps would import the full Commodity Exchange Act apparatus, including anti-fraud and anti-manipulation authority under Section 6(c)(1) and Rule 180.1, large-trader reporting, and registered intermediaries. It would also strengthen the argument that federal law preempts conflicting state gambling rules, and it would give institutions a familiar, regulated instrument to trade.

Do AI bots really dominate Polymarket and Kalshi?

On Polymarket, more than 30% of active wallets are autonomous agents and 14 of the top 20 accounts by volume are bots, with roughly 37% of agents booking a profit against an estimated 7% to 13% of humans. Winning is still hard, however: studies show that trade direction inferred from the public feed matches the on-chain truth only about 59% of the time, and naive off-the-shelf models tend to perform no better than a coin flip.

Will the Supreme Court decide whether Kalshi’s contracts are swaps?

It might. New Jersey has petitioned the Court, and an unusually broad coalition has urged it to take the case, including 39 states and the District of Columbia, the NFL, tribal groups, the American Gaming Association, and the original architects of the Dodd-Frank Act. Kalshi’s response is due November 9, 2026, and the Justices have not yet agreed to hear it; a three-way circuit split makes review plausible but far from certain.

By Marcus Okafor, senior markets correspondent at HOGE Wire.