A sports contract has a finite set of people who can move the outcome or know it early. They are on a team sheet, an officials roster, a tournament entry list. The exchange can name them, and a court can check the naming. An election contract has no such list. According to the 2024 Election Administration and Voting Survey published by the U.S. Election Assistance Commission, the 2024 general election was run by 6,461 local jurisdictions and staffed by 675,434 poll workers on Election Day, with a further 149,321 working early voting. That is the resolution source. It is also, on the plain wording of the rules that already exist, a very large population of restricted participants that nobody has written down.
This is not a gap in the rulebook. The rulebook is wide enough. It is a gap between a prohibition and the record that would let anyone enforce it, and the months since August 2026 have produced an unusual amount of evidence about what happens when that gap is left open.
The prohibition is already broad, and it already reaches the county clerk
The relevant text is not hard to find. KalshiEX LLC filed version 1.18 of its rulebook with the CFTC in July 2025, and Rule 5.17 contains two separate prohibitions that bear on this. Paragraph (y) defines an Insider as any person who has access, or is in a position to have access, to material nonpublic information before it is made public, and then adds a second category on its own terms: a trader who is an employee or affiliate of a Source Agency for a contract is barred from trading that contract. Paragraph (z) is broader still. It bars anyone who is a decision maker, directly or indirectly, or who has any influence on the outcome, in the rulebook's own words, "no matter the scale and importance of the influence".
The rulebook defines Source Agency as the agency that publishes the underlying or the expiration value for a contract. For an election contract, that is the body that certifies the result. In the United States, that body is not one agency. It is thousands of county and municipal offices, each with its own staff, its own seasonal workforce and its own certification timetable.
So the reach of paragraph (y) is not in doubt. A ballot-processing clerk in one of those 6,461 jurisdictions is an employee of a Source Agency. What is in doubt is whether any exchange holds a list that would let it tell that clerk apart from any other new account. The rulebook itself hints at how the mechanism is meant to work: it defines a Source Agency Prohibition as the list of Source Agencies that may implicate trading prohibitions. The list is of agencies, not of people. For a contract whose Source Agency is a single federal statistical office, that distinction costs nothing. For a contract whose Source Agency is the American election system, it is the whole problem.
The candidate was findable and the count clerk is not
August 2026 supplied a clean demonstration of where the current tooling works. On 31 August 2026 Kalshi announced a disciplinary action against Laurie Buckhout, the Republican nominee in North Carolina's first congressional district, for trading contracts tied to her own race. According to WRAL reporter Will Doran, who reported the action the same day, the resolution was a three-year suspension and a financial penalty of 2,589.96 dollars. Buckhout told WRAL: "I bet on myself. Literally." She described it as a mistake and said she had worked to resolve it once she learned there was a problem.
Read as a compliance story rather than a political one, the case is encouraging. A candidate is a named, published, verifiable person. Ballot access filings make the list for you. Rule 5.17(z) applies to a candidate without any judgement call, because a candidate's ability to influence whether she wins is influence by definition. Surveillance had something to match against, and the match was made.
Now change one variable. Replace the candidate with a temporary worker who handles ballots in a county of forty thousand people, opens an account under her own name, and buys the side she has reason to think is correct three hours before the numbers are posted. Every element of Rule 5.17(y) is satisfied. Nothing in the exchange's data identifies her. There is no filing, no roster, no public register of the several hundred thousand people who hold that role for a few weeks every two years.
This is the same distinction we drew in our earlier piece on inside information, but the election case sharpens it. The problem is not that the prohibition is too narrow. It is that enforceability depends on an identification layer, and that layer is built one industry at a time.
Three sets of restrictions arrived this year and none of them reach the counting room
The interesting development of 2026 is that the restrictions did start arriving. They simply arrived from the direction of the source, not from the direction of the exchange, and they stop short of the people who actually produce the number.
The first is federal and it is binding. On 30 April 2026 the Senate agreed to S. Res. 708, submitted by Senator Bernie Moreno, which amends Rule XXXVII of the Standing Rules of the Senate. The new paragraph 15 provides that no Member, officer or employee of the Senate may enter into an agreement or transaction for an excluded commodity, as defined in section 1a of the Commodity Exchange Act, that depends on the occurrence or non-occurrence of a specific event. The only carve-out is insurance in which the insured holds a lawful insurable interest. Section 2 records the sense of the Senate that the House, the executive branch and the judicial branch should adopt similar restrictions.
The second is proposed and broader. H.R. 7004, the Public Integrity in Financial Prediction Markets Act of 2026, was introduced on 9 January 2026 by Representative Ritchie Torres with a long list of cosponsors. It would make it unlawful for a covered individual to knowingly trade a prediction market contract related to government policy, government action or a political outcome while holding relevant material nonpublic information, or where that information may reasonably be obtained in the course of official duties. The definition of covered individual is the part worth reading closely. It lists elected officials of the federal government, employees of the House or Senate, political appointees, and employees of an Executive agency as defined in section 105 of title 5. State officials are not there. County officials are not there. Poll workers are not there.
The third set is local, voluntary and the only one aimed at the counting room itself. Votebeat editor-in-chief Carrie Levine reported on 10 August 2026 that Delaware County, Pennsylvania elections director Jim Allen had added an explicit ban on poll workers using prediction markets, comparing the practice to referees betting on a basketball game. By September the pattern had spread. CNBC reported on 11 September 2026 that Delaware County's policy covers roughly 2,200 employees, and Arizona's Family reported on 16 July 2026 that Arizona Secretary of State Adrian Fontes had adopted a policy barring his staff from trading on elections or election-related events, that Maricopa County had approved a comparable measure, and that Governor Katie Hobbs had signed an executive order addressing the use of confidential information for profit on these markets.
Put the three together and the shape is clear. One legislature has bound itself. One bill would bind the federal executive. A handful of counties and one state office have bound their own staff. Between them they do not cover a meaningful fraction of 6,461 jurisdictions, and there is no mechanism by which they ever would, because each one is a separate employer making a separate decision.
Wisconsin answered a different question, and the answer points the other way
One state response is worth separating out, because it is often filed alongside the staff policies and it is not the same thing at all.
In July 2026 the Wisconsin Elections Commission issued guidance warning voters that trading election contracts may fall foul of state law. The statute it relies on is old and short. Wisconsin Statutes section 6.03(2) provides that no person shall be allowed to vote in any election in which that person has made or become interested, directly or indirectly, in any bet or wager depending on the result of the election. Reporting carried by the Wisconsin Law Journal on 21 July 2026 noted that the commission named both Kalshi and Polymarket, and that knowingly voting while disqualified is a Class I felony carrying a maximum of a 10,000 dollar fine and three and a half years.
Notice the target. Wisconsin's rule restricts the voter, not the administrator. It is a disenfranchisement provision from a different century, aimed at the idea that a person with money on a race should not also cast a ballot in it. It does nothing about the clerk who counts the ballot, and a venue that treated the Wisconsin guidance as evidence that election integrity is being handled by the states would be reading it backwards.
Why this is a settlement problem and not only an integrity problem
It is tempting to file all of this under insider trading and move on. That undersells the cost, and the officials quoted in the Votebeat reporting were describing something else entirely.
Dean Logan, the Los Angeles County Registrar-Recorder and County Clerk, told Votebeat that when early returns diverged from where a market had been trading, the divergence itself generated suspicion about the way ballots were being processed. He put it in terms of perception and misinformation rather than trading. That is a settlement-source problem in the precise sense we set out in our article on settlement risk: a price that is public, continuous and confident is being read as a claim about what the count should produce, by an audience that never agreed to the contract terms and has no reason to know how thin the market was.
A fragile settlement source is usually described as one that might report late, report wrong, or stop reporting. The election case adds a fourth failure mode that the category has not priced. The source can be pressured by the existence of the market. Nothing about the count changed. What changed is that a number now exists against which the count is publicly compared in real time, and the people doing the counting have to answer for the gap.
That is not an argument for delisting election contracts. We have argued the opposite: these markets have real informational value, and the classification question around them is genuinely unsettled, as we set out in our piece on the contest exclusion. It is an argument that a venue listing them takes on an obligation that does not arise with a contract settled by a single published index.
What a venue should be able to produce on this, today
Surveillance is not a courtesy in this category, it is a requirement. 17 CFR 38.156 obliges a designated contract market to maintain an automated trade surveillance system, and 38.157 requires real-time monitoring. Neither tells an exchange how to identify a population it has no roster for. That design work belongs to the venue, and it is the kind of thing Directive 02 asks to be stated in plain language on the surface where the trade happens.
Four things are answerable now, without waiting for a rulemaking.
- Say which Source Agency settles the contract, by name, in the contract terms. For an election contract that means naming the certifying authority for that specific race, not "official results". A trader cannot assess a fragile source that is never identified, and a compliance team cannot scope a restricted population it has never written down.
- State what the trading prohibition actually covers for this contract type. Rule 5.17(y) already extends to employees and affiliates of a Source Agency. If that is understood to include seasonal poll workers in 6,461 jurisdictions, say so at account opening and at the point of trade, in words a temporary worker would recognise as applying to her.
- Publish the attestation, and publish what happens next. An attestation at signup is not a roster, but it is a record, and it converts an unenforceable prohibition into a verifiable false statement. What matters is whether anything is done with it: sampling, escalation on unusual timing in county-level contracts, and a published account of how many accounts were restricted.
- Describe the detection method without claiming more than it delivers. Kalshi executive Sara Lang told WRAL the exchange has "robust protection, security, surveillance in place to make sure that it isn't happening". That is a strong claim, and the Buckhout case shows it working on the one category of person a public filing already identifies. The honest version distinguishes between populations the venue can identify and populations it cannot, and says which is which.
None of this requires a regulator to act first. All of it is the kind of thing a venue would rather have written down before someone asks it under oath, which is the argument Directive 05 makes about setting standards against the jurisdiction that regulates next.
The list is the product, not the paperwork
The category has spent 2026 arguing about classification: whether an event contract is gaming, whether it is a swap, which authority decides. Those fights matter. But the Buckhout case and the county policies point at something smaller and more immediate, and it is not waiting on any of the litigation.
A restricted participant list is not compliance overhead. It is the mechanism by which an exchange converts a rule it has already written into something that can be enforced, audited and shown to a sceptical outsider. Where that list is short and published, as it is in sport, the rule works. Where the list would have to contain 675,434 people that nobody has ever enumerated, the rule exists and does nothing, and the venue's own rulebook is the document that says so.
The honest position, as of 22 September 2026, is that no venue can produce that list, and that the restrictions which did arrive this year came from the Senate, from one state office and from a handful of county elections directors, rather than from the exchanges holding the contracts. That order tells you something. The floor is being set by the people who run the source, one employer at a time, while the market that depends on the source watches it happen. Our view is that a category which wants to be trusted with contracts of this kind should be the one writing that floor down first.