Summary
- Polymarket operates private election-market detection while public authorities and lawmakers have not defined common rules for election-market conduct.
- Shana Bautista said Polymarket traces funds, investigates networks, evaluates legitimate explanations and feeds case findings into new detections, but withholds specific criteria and methods.
- NPR reported that at least two Polymarket investigations were sent to the CFTC, leaving the agency’s procedural treatment and outcome unresolved.
- Rajiv Sethi told NPR that election prediction markets can shape campaign momentum and become self-reinforcing, raising a risk distinct from private detection.
- Kalshi’s April action against three candidates shows platform-imposed consequences while leaving comparative surveillance quality and independent review unestablished.
An Oct. 27, 2026 report describes Polymarket operating an election-surveillance system and investigating election-market bets ahead of the Nov. 3 midterms. Shana Bautista, Polymarket’s head of Global Intelligence & Investigations and a former FBI official and cryptocurrency analyst, told NPR that her team traces funds, investigates networks, evaluates legitimate explanations and documents the basis for its decisions. NPR reported that at least two investigations were sent to the U.S. Commodity Futures Trading Commission, although the report does not establish what followed those referrals. Bautista declined to disclose the detection criteria or methods used. The report also said Kalshi banned and fined three political candidates in April for betting on themselves, while regulators and lawmakers pursued controls on increasingly popular prediction markets. The institutional consequence is a sequencing problem: private detection is operating while public accountability remains unsettled. Rajiv Sethi, an economics professor at Barnard College at Columbia University, identified a separate risk by telling NPR that election prediction markets may influence campaigns and become self-reinforcing. The available evidence establishes the reported architecture of Polymarket’s surveillance system, but not its accuracy, sufficiency or effectiveness.
What the reported system does
Bautista said Polymarket builds detections around how someone would try to manipulate an election market and what that activity would leave behind. She stated in a company statement:
“We build detections around how someone would actually try to manipulate an election market and what that activity would leave behind.”
Her team’s approach, according to Bautista, combines funds tracing with investigation of sophisticated networks. The team connects evidence, evaluates legitimate explanations and documents the basis for decisions. Bautista said trading on protected information violates platform terms of use and the law. She stated the standard as follows: “If you have information that you have a duty to protect, don’t trade on it,” adding that doing so was “explicitly against our terms of use on both platforms, but also is against the law.”
That legal proposition remains Bautista’s attributed position. The supplied report identifies no statute, regulation, rule or adjudication independently establishing it.
In a hypothetical example, Bautista described investigators recognizing a possible attempt to materially move an election market through several accounts trading in short bursts, consistent device identifiers across transactions, and new wallets associated with four or five accounts. She said findings from completed cases feed back into the process for creating future detections. The company’s stated method therefore combines observation of behavioral patterns with a continuing internal feedback loop.
Bautista said insider trading and market manipulation look the same regardless of the market. Her team is looking for threat vectors and risk areas that have existed in traditional markets for decades, while recognizing that established tactics are being applied through prediction markets. The article does not report how many investigations Polymarket opened, how many it resolved, what criteria triggered scrutiny or whether the system has produced false positives or missed conduct.
The company’s detection system also does not complete the enforcement process on its own. NPR reported that at least two investigations Polymarket said it had opened were sent to the CFTC. The report does not establish whether the CFTC opened, accepted jurisdiction over, adjudicated or resolved those matters.
Kalshi’s April action provides a different kind of record. The supplied report says Kalshi banned and fined three political candidates for betting on themselves; AP and PBS excerpts cited during review describe the action as suspending and fining the candidates. That action shows a platform-imposed consequence, but it does not establish the legality, accuracy or independent review of surveillance at either venue.
Sethi told NPR that election prediction markets are uniquely vulnerable to manipulation and protected-information trading. He said such markets “have the possibility of being a self-fulfilling prophecy,” because beliefs about a candidate’s momentum can affect the campaign. The surveillance system described by Bautista addresses one part of the information environment: conduct within election markets. The dynamic described by Sethi concerns how market prices may influence conduct outside them.
Who bears the burden, and what remains unverified
The regulatory burden falls on platforms, political candidates, traders, public agencies and voters. Voter confidence can determine whether election markets retain public legitimacy as an information-setting institution.
The source provides substantial detail about Polymarket’s surveillance system and one documented Kalshi enforcement action. Equivalent detail is absent from Kalshi, the CFTC and lawmakers. Information is also largely absent from candidates, traders, election administrators and voters.
That imbalance prevents a comparison of enforcement quality. The record does not establish whether the platforms use similar thresholds, treat cases consistently, notify candidates or provide independent review of adverse determinations.
The same independence standard applies to both companies. Neither platform’s enforcement record can serve as an independent assessment of its own performance. Kalshi’s candidate sanctions show that a platform imposed consequences; they do not establish that Kalshi’s detection system is more accurate, that its rules are legally settled or that either platform subjects its determinations to review independent of the company investigating the matter.
The evidence also supports a distinction between public uncertainty and demonstrated failure. No source establishes that Polymarket’s system is working as represented, and no source establishes that it is failing. The established fact is narrower: according to Bautista, the company has deployed surveillance and referred investigations while withholding operational details.
The policy design space
The available policy options address different parts of the enforcement chain and carry different institutional burdens.
| Option | Design | Principal benefit | Principal exposure or limitation |
|---|---|---|---|
| Platform-led enforcement | Operators apply private terms, monitor activity and make selected referrals | Fast adaptation | Different standards and disclosure practices across platforms |
| Statutory control | Laws define conduct, jurisdiction and institutional responsibilities | Common baseline | Implementation lag |
| Hybrid | Confidential detection criteria plus published duties, aggregate reporting, notice and appeal, and regulator review | Accountability without operational disclosure | Requires a functioning independent-review channel |
| Restriction | Limit political contracts while rulemaking proceeds | Boundary case for near-term risk | Potential loss of the market’s price-information function |
| Candidate prohibition | Prohibit candidates from trading in markets connected to their own campaigns or races | Easily communicated rule | Scope, exceptions and cross-venue consistency |
| Joint platform-regulator system | Operators and public authorities share procedures, evidence standards and case reporting | Greater consistency and investigative scale | Coordination, privacy and institutional authority |
| Defer and monitor | Authorities collect cases while platforms refine controls | Flexibility | Fragmentation may persist |
| Market-design review | Examine market design, information integrity and public reporting rather than focusing only on individual conduct | Broader treatment of election information | May not resolve which conduct should be prohibited |
The options are not mutually exclusive. A combined arrangement could establish a public conduct standard while retaining confidential detection methods, requiring aggregate reporting and preserving secure referral channels. The available evidence does not establish that such a combined framework has been adopted.
The unresolved legal proposition advanced by Bautista also limits what the numerical comparison can establish. A legal-consistency filter applies only if authoritative sources confirm that protected-information trading is prohibited by independently established law; the numerical options do not themselves legitimize conduct. Likewise, statutory surveillance standards discussed in the comparison below are analyst-designed options, not features attributed to the reported legislative proposals.
Decision criteria
A complete architecture must distinguish conduct rules, detection, referral, adjudication, sanctions and public reporting. A policy focused only on sanctions could leave detection and referral unchanged, while a surveillance-only policy could leave inconsistent platform rules intact.
Seven criteria distinguish the options:
- Outcome quality: Whether protected-information trading and coordinated manipulation are detected, whether behavior is deterred, and whether election markets remain useful indicators rather than self-reinforcing signals.
- Cost and capacity: Whether surveillance, funds tracing, network analysis and case processing have sufficient institutional resources to support enforcement.
- Risk: Whether private screening produces false positives, categorical rules discourage legitimate activity, and monitoring raises questions about retention and use of device, wallet and trading data.
- Institutional fit: Which body sets rules, which operator investigates, which authority receives cases, and whether similar conduct receives similar treatment across venues.
- Transparency and accountability: Whether aggregate rules and outcomes can make enforcement explainable without disclosing thresholds that could assist evasion.
- Reversibility: Whether a policy can adapt as tactics change, compared with a statutory framework that may be harder to amend when written around evolving techniques.
Outcome quality and deterrence overlap. Transparency and accountability are related but distinct: the former concerns whether misconduct is prevented, while the latter concerns whether the basis for enforcement can be reviewed.
The article supplies no comparable data on cost, false-positive rates, investigative performance, case duration or the number of market participants affected. It therefore does not support an evidence-based weighted ranking or reliable assignment of probabilities to the alternatives.
The disclosed multi-criteria comparison
The following ELECTRE II-style model uses analyst judgments rather than measured effects. Scores are displayed from 0 to 100, tied criteria receive no concordance credit, and an alternative outranks another when concordance exceeds (C^=0.50) while discordance falls below (D^=0.25).
[ C_{ab}=\sum_{a_j>b_j}w_j ]
[ D_{ab}= \frac{\sum_{a_j<b_j}w_j|a_j-b_j|} {\sum_{a_j\ne b_j}w_j|a_j-b_j|} ]
The four compared postures are the current platform-led posture, statutory control, a hybrid arrangement and restriction of political contracts.
| Criterion (weight) | A Current | B Statutory control | C Hybrid | D Restriction |
|---|---|---|---|---|
| Speed of effect before Nov. 3 (.20) | 80 | 20 | 60 | 75 |
| Marginal deterrence of manipulation and protected-information trading (.20) | 45 | 55 | 65 | 40 |
| Protection of legitimate traders from misdirected enforcement (.15) | 50 | 55 | 70 | 70 |
| Oversight information value (.15) | 30 | 75 | 80 | 55 |
| Compatibility with confidential detection criteria (.10) | 95 | 40 | 85 | 90 |
| Independence of oversight from platform cooperation (.10) | 25 | 90 | 60 | 70 |
| Preservation of the market’s price-information function (.10) | 90 | 70 | 85 | 20 |
The hybrid outranks the current posture: concordance is (0.60), and discordance is (5.5/(5.5+18.0)=0.234). It also outranks statutory control: concordance is (0.90), and discordance is (3.0/(3.0+19.0)=0.136).
The hybrid does not outrank restriction. Concordance is (0.45), comprising (0.20) for deterrence, (0.15) for oversight information value and (0.10) for price-information preservation; protection of legitimate traders is tied and receives no credit. Discordance is (4.5/(4.5+15.25)=0.228). Because (0.45<0.50), neither threshold condition is met. In the reverse comparison, restriction has concordance (0.40) and discordance (15.25/(15.25+4.5)=0.772), so restriction does not outrank the hybrid.
The remaining directional comparisons are:
- Current posture versus statutory control: (0.40/0.451), with the reverse at (0.60/0.549).
- Current posture versus restriction: (0.60/0.542), with the reverse at (0.40/0.458).
- Statutory control versus restriction: (0.55/0.584), with the reverse at (0.45/0.416).
Those displayed figures show statutory control outranking the current posture, the current posture outranking restriction, and statutory control outranking restriction. Describing the hybrid as the only alternative with any outranking relation would therefore misstate the model.
The hybrid’s relation to restriction remains unresolved. Its selection as the immediate design rests on a disclosed qualitative tie-break rather than formal domination of every alternative.
The primary sensitivity in the hybrid–current-posture comparison is the current posture’s deterrence score. If that score rose from 45 to 65, the alternatives would tie on that criterion; hybrid concordance would fall from (0.60) to (0.40), and the outranking relation would fail.
If statutory control’s independence score fell from 90 to 60, hybrid concordance would remain (0.90), because the hybrid would win the other six criteria. Hybrid discordance would fall to zero because independence would be tied. Statutory control would still score materially higher on oversight independence: the scalar model allows gains on several other criteria to offset that single disadvantage, so the numerical result does not make the hybrid independent.
Independent regulator access remains a substantive condition for a statutory alternative, while an independent-review channel remains necessary for the hybrid’s immediate selection.
How the system could fail
The most consequential failure would be an election-integrity episode involving material conduct accompanied by a public record unable to explain how the conduct was detected, what action followed or whether the system performed as represented.
Three present decision points raise that risk:
- Withholding detailed criteria without publishing aggregate accountability measures.
- Treating a referral as the end of the enforcement record rather than the beginning of documented external review.
- Deferring legal-authority and oversight-independence questions until after the election.
Observable warning signs need not reveal detection methods. They include continued enforcement language without aggregate case data, referral language without a stated procedural outcome, and continued absence of a published protected-information standard on either platform.
Withholding criteria leaves outsiders unable to test whether a signal is probative, whether legitimate explanations were adequately considered or whether similar conduct would be treated consistently. The public record does not state why the criteria are withheld, and it does not establish that withholding caused any particular omission.
The internal feedback loop can refine detection, but it concentrates information inside the platform. Without external review of case files or resulting criteria, the public may not distinguish adaptation to new tactics from repetition of established tactics.
Continued referrals can show that activity is occurring. They do not establish detection accuracy, the absence of missed tactics or the disposition of referred matters. Aggregate reporting would therefore function as a diagnostic instrument rather than merely an informational addition.
Public-law uniformity and operational capacity
Public-law uniformity and operational enforcement capacity are separate axes. A statute can be enacted before agencies develop the staff, procedures or data systems needed to apply it, while platform surveillance and referral practices can improve without new legislation. Their intersection determines whether rules are both common and enforceable.
| Lower operational capacity | Higher operational capacity | |
|---|---|---|
| Lower public-law uniformity | Private patchwork: Platforms retain primary enforcement responsibility. Indicators: legislative proposals fail to advance, public case outcomes remain limited and platform rules diverge. Posture: robust minimum safeguards with contingent review if referrals rise without matching public capacity. | Coordinated enforcement without comprehensive legislation: Operators or agencies exchange evidence and apply comparable remedies. Indicators: cross-platform referrals increase, sanctions become more consistent and coordination procedures become public. Posture: a scenario-dependent compact that may later be codified. |
| Higher public-law uniformity | Rules ahead of enforcement: Public obligations exist, but guidance, staffing, data standards or case processing lag. Indicators: laws or guidance are adopted without timely outcomes, or referral volumes rise while public resolutions remain slow. Posture: capacity-building before imposing additional obligations. | Integrated public-private regime: Public authorities define responsibilities, operators detect and cases move through predictable referral and review processes. Indicators: jurisdiction is clear, joint reporting is regular and comparable outcomes follow across platforms. Posture: broadly robust with periodic recalibration. |
Forces, dependencies and the stress case
Social and institutional forces include the possibility that election markets influence campaign momentum. Technological forces include funds tracing, device-level analysis, wallet tracing and adaptive detection systems. Economic forces arise from attempts to materially move an election market. Political forces include the CFTC referrals, Kalshi’s candidate sanctions and lawmakers’ pursuit of legislation.
The report identifies no environmental driver. Environmental disruption remains a residual scenario input rather than a load-bearing assumption.
Several safeguards do not depend on a single prediction about legislation or enforcement capacity. Evidence preservation, secure referral channels, cross-platform coordination and aggregate public reporting can preserve confidential methods while exposing the scope of monitored conduct and reported results.
Other institutional choices carry dependencies. A comprehensive legislative framework depends on enforcement capacity following the rules. A categorical candidate prohibition depends on consistent adoption and review across venues. A joint platform-regulator system depends on cooperation, information-sharing authority and sufficient resources.
Capacity-building becomes more urgent if rules precede capacity. Stable operating practices can be converted into standards if enforcement improves first. Increased case volume would call for clearer public reporting, and significant cross-market events would raise the value of rapid-response rules.
An unobserved stress case is a coordinated cross-platform event severe enough to overwhelm current private monitoring and trigger emergency intervention, moving legal authority and enforcement capacity together. Its probability is unknown, it is not a reported event, and the decision not to disclose detection methods prevents the article from establishing preparedness for it.
Detection and oversight scenarios
The report supports several present conditions: the election-surveillance system is operating according to Bautista’s account; her presently disclosed position is that detailed criteria and methods will not be discussed; at least two investigations were sent to the CFTC; and lawmakers were seeking legal controls.
Continued nondisclosure is a working assumption that could change if a platform or regulator adopted a different policy. The first consequential uncertainty is whether detection remains ahead of adaptation as the feedback loop incorporates case findings. The second is whether external oversight arrives early enough to affect the election cycle.
| Detection performance | Oversight timing | Likely indicators and interpretation |
|---|---|---|
| Detection holds | Oversight arrives in time | Aggregate investigation and outcome figures; a regulator framework or audit discussion; published alignment between platform protected-information rules; evidence that referred cases produce documented external review. |
| Detection holds | Oversight delayed | Investigations and referrals continue while performance measures remain unavailable; legislative inquiries may state that more information is needed; platform statements may continue without case data. The system may be functioning, but the public cannot verify that it is. |
| Detection erodes | Oversight arrives in time | A regulator may issue guidance or begin a proceeding while platforms provide little performance information; investigators may close matters without action or emphasize enforcement without publishing outcomes. Timely oversight can reveal the gap only if it receives information independent of the platforms. |
| Detection erodes | Oversight delayed | Performance and outcome data remain unavailable while referral activity may continue, contract or cease. The absence of published detection measures prevents an observer from determining whether the process is working or idle. |
The delayed-oversight states cannot be distinguished from detection-performance and case-outcome information alone. They can still differ through referral activity and platform statements. Aggregate reporting is therefore necessary to distinguish a functioning but unverifiable system from an idle or deteriorating one.
The governance sequence
Detailed operational secrecy and public accountability are separable. The record supports keeping detection criteria out of public view while requiring a published interface of duties, aggregate case data, notice and appeal rights, and external review.
Within the disclosed model, the hybrid is the immediate design because it outranks the current posture and statutory control while remaining compatible with confidential detection criteria. Its relation to restriction is unresolved, so that selection includes a disclosed qualitative tie-break. The selection depends on retaining an independent-review channel; confidential detection alone does not provide independent oversight.
Before the Nov. 3 election, the disclosed assessment calls for Polymarket and Kalshi to publish their governing protected-information standards, report aggregate numbers of investigations opened, actions taken and traders cleared, and provide written notice and an appeal path for challenged conduct. Detailed detection methods need not be disclosed for those accountability measures to be verifiable.
Immediately after the election cycle, a referral should lead to a recorded external review rather than terminate at a platform-generated count. Any statutory design containing surveillance standards is an analyst-proposed addition to the reported effort to enact laws controlling prediction markets, not a feature the article attributes to lawmakers.
Beyond the election cycle, oversight requires regulator audit and reporting authority that does not depend on the cooperation of the platform being examined.
Both a hybrid regime and an enacted statutory regime can place evidence outside the platform, but their timing differs. The hybrid supplies an interim review channel, while a statutory regime with independent audit authority represents the more durable endpoint. The statutory endpoint is slower than platform-led intervention. That delay is necessary to avoid leaving the evaluated party as the sole provider of the evidence used to assess it.
The supplied record does not establish how well surveillance works. The governance sequence can make the system assessable without publishing operational methods, but its efficacy cannot yet be evaluated.
The limits of the public record
The report frames private surveillance in operational terms—detections, tracing, network investigation, legitimate-explanation review and feedback—but supplies no basis for describing the system as accurate, complete or effective.
The public record contains company statements, the report that at least two matters were sent to the CFTC, and Kalshi’s April enforcement action. The unresolved gap concerns detection performance, case outcomes and the basis for decisions.
No counts are provided for investigations opened, cases resolved, positions cleared, false-positive allegations or detection misses. No case description connects the reported CFTC referrals to particular conduct, and no error or performance measures are published.
Bautista’s stated method includes evaluating legitimate explanations, placing error control inside the investigative process. No data permit assessment of how often errors occur. The detailed criteria withheld by Polymarket cannot be assessed for probative value, consistency or evasion risk.
A referral to the CFTC establishes a transfer of cases, not agency acceptance, jurisdiction, an opened proceeding or an outcome. Kalshi’s action demonstrates a platform consequence but not comparative surveillance quality, legal authority or independent review.
The evidence does not establish that private surveillance is sufficient, that Polymarket and Kalshi are converging, or that any proposed statute would resolve jurisdictional uncertainty. The supplied evidence supports only Bautista’s reported account of the surveillance system and the Nov. 3 timing; independent corroboration of the precise deployment and date was unavailable in the reviewed evidence.
The supportable conclusion is bounded: platforms are operationalizing detection while public authorities consider broader controls, but the public cannot yet determine whether the system is functioning as represented.
Evidence that could change the assessment
- Published jurisdictional rules.
- Aggregate cross-platform enforcement data.
- Disclosure of candidate notice and review procedures.
- Evidence about detection accuracy, false-positive rates and institutional capacity.
- Aggregate investigation and outcome figures.
- The CFTC’s procedural treatment of the referred matters.
- A regulator framework, audit discussion or documented external-review channel.
- Published alignment between platform protected-information standards.
- Authoritative statutory, regulatory or judicial grounding for the legal proposition advanced by Bautista.
Analytical techniques used in this piece
This analysis applies the methods below. Each links to a short, plain-English explainer you can read and reuse.
- Decision Architecture
- Designs the structure of a high-stakes decision — sequencing, gates, and what to settle first.
- Multi-Criteria Decision Analysis
- Scores competing options against several weighted criteria at once.
- Scenario Planning
- Builds a small set of distinct, plausible futures to plan against.