What the 2026 Problem-Gambling Prevalence Data Actually Tells Regulators — and Why Enforcement Is Tightening

Three regulators published material problem-gambling data sets in the twelve months to mid-2026. Each one immediately translated into enforcement pressure. The Autorité Nationale des Jeux (ANJ) in France deployed an algorithmic detection tool that identified a problem-gambling population seven times larger than operators had self-reported, then used the resulting data gap to anchor a €500,000 fine. The UK Gambling Commission’s Gambling Survey for Great Britain (GSGB) recorded a PGSI 8+ rate of 2.4% alongside product-level harm data that is now driving the architecture of financial risk assessments. In Sweden, Spelinspektionen’s own finding that self-excluded players are the primary users of the unlicensed market produced two binding compliance instruments in 2026: a comprehensive credit ban effective 1 May and new Spelpaus API requirements under SIFS 2026:3 effective 1 August. Compliance officers need to understand these data sets not as public-health statistics but as the evidentiary foundation for the next supervisory wave.

The ANJ Algorithm: A New Compliance Benchmark for Problem-Gambler Detection

The ANJ developed its detection algorithm using 23 indicators drawn from scientific literature, applied to continuous account-level gambling data submitted by licensed operators. The indicators cover financial transaction patterns, frequency and intensity of gambling activity, use of moderation tools, self-exclusion history, and betting history over time. Players are classified into four groups: recreational, moderate, excessive, and manifestly excessive.

Applied to H2 2025 data, the algorithm identified approximately 600,000 players with a high probability of excessive gambling. That figure represents 8.7% of the total online account-based gambling population across French-licensed operators, including accounts at FDJ United and Pari-Mutuel Urbain (PMU). Of the 600,000, around 300,000 were classified as “manifestly excessive,” a category the ANJ described as requiring operator identification as a matter of urgency. These 600,000 players generated approximately €1.2 billion in GGR, accounting for 60% of total online market revenue in France.

“The upward trend since 2023 was likely due to overall market growth, but does not explain the whole picture as the number of problem gamblers has risen faster than the total number of gamblers.” The ANJ made this observation in its May 2026 algorithm publication, warning that the divergence between market growth and harm growth demands a structural operator response.

The stark data divergence is the compliance signal. Prior to the algorithm’s publication, operators were collectively reporting approximately 89,000 problem gamblers, a figure that had itself tripled between 2024 and 2025 under growing regulatory pressure. The algorithm’s estimate of 600,000 means the operator-identified population was running at roughly 15% of the regulator’s estimate. The ANJ described this discrepancy as “inconsistent with the size of the operators’ player base and prevalence studies.” That framing converts a methodological disagreement into an enforcement basis: if the regulator can demonstrate that your detection rate diverges materially from a validated population model, your compliance controls are evidentially deficient.

The ANJ’s 2024, 2026 strategic plan had already established reducing excessive gambling as a regulatory priority. Problem gambling prevalence among French online gamblers as a whole stands at 8.7%; among sports bettors specifically, the figure reaches 15.3%. The French Monitoring Centre for Drugs and Drug Addiction estimated in 2024 that approximately 1.17 million people in France exhibit problematic gambling behaviour, with around 360,000 classified as excessive. The algorithm is designed to close the gap between those population-level estimates and what licensed operators are actually detecting in real-time account data.

ANJ Decision 2026-031: When Prevalence Data Becomes an Enforcement Instrument

The connection between the algorithm’s findings and direct enforcement became concrete on 10 July 2026, when the ANJ published Decision n°2026-031, imposing a €500,000 fine on an unnamed online betting operator referred to in the decision as Company X. The investigation covered the period 1 October 2023 to 31 March 2024, during which the ANJ conducted an administrative inquiry into Company X’s account data stored in the operator’s secured data vault.

Using a scoring system based on deposit frequency, betting intensity, loss patterns, and self-exclusion history, the regulator identified the 30 accounts with the highest risk profiles across Company X’s player base. Of those 30, 29 were the subject of formal grievances. Six players were missed entirely by the operator’s own systems, 23 were misclassified at a lower risk tier. Company X also failed to provide proportionate, graduated interventions to 25 of those players to moderate their gambling behaviour. The combined net losses recorded for the 29 players during the review period were material, and the ANJ noted that Company X’s practice of offering bonuses to players already exhibiting high-risk indicators was itself an aggravating factor.

The ANJ’s decision rejected Company X’s defence, which challenged the legal definitions of “identification” and “accompaniment” as insufficiently precise, and upheld the authority of the 2021 ANJ reference framework for operator obligations. Critically, the decision established that “identification” and “accompaniment” are distinct and independently enforceable duties. An operator that has identified a high-risk player but provided inadequate graduated support commits a separate violation from one that failed to identify the player at all.

What Does the GSGB Actually Tell the UKGC?

The 2025 Gambling Survey for Great Britain (GSGB), published across Waves 3 and 4, recorded a PGSI score-8-or-above rate of 2.4% of adults, a modest decline from 2.7% in 2024. Moderate-risk gambling (PGSI 3, 7) increased slightly. Overall gambling participation dipped from 60% to 59%, and online gambling participation held steady at 38%. The Gambling Commission extended the GSGB’s delivery contract in November 2025, awarding a four-year agreement running from 2026 to 2029 to the National Centre for Social Research and the University of Glasgow, confirming the survey as the Commission’s primary ongoing data infrastructure.

Industry critics have questioned the GSGB’s methodology, arguing that the survey suffers from topic salience bias and over-samples engaged gamblers. The Gambling Commission has separately acknowledged the challenge of coherence between GSGB data and operator-level industry data. A coherence review with the Bingo Association was published in April 2026, and the Commission’s Data Innovation Hub is tracking the same issue for illegal gambling volumes. These methodological debates are legitimate, and compliance officers should understand them, because they inform how the UKGC uses its data: not as a precise prevalence count, but as a risk-stratification tool.

The product-level data within the GSGB is more operationally significant than the headline PGSI figure. The October 2025 GSGB publication showed that online slots carry a higher-than-average proportion of players with a PGSI score of 8 or more. That finding is reproduced in the UK government’s 2026 consultation response on the tax treatment of remote gambling, which explicitly cited GSGB data alongside Health Survey for England figures. The government document noted that the Health Survey for England estimated 0.3% of adults have a problem with gambling and 2.8% are at at-risk levels, and that NHS referrals for gambling addiction increased by almost 91% in the year to December 2024, with NHS England doubling its network of specialist clinics.

The operational consequence of the online-slots harm signal is the financial risk assessment (FRA) framework. The proposed architecture uses two thresholds. A “light check” triggers when a customer loses more than £125 over 30 days or £500 in a year, using publicly available data such as bankruptcy records. An enhanced risk assessment triggers when losses exceed £1,000 within 24 hours or £2,000 over 90 days, using credit reference data to identify signs of serious financial distress. The UKGC’s stated objective is for 97% of checks to complete automatically, with fewer than 3% of active accounts triggering any form of direct intervention. UKGC Executive Director Tim Miller clarified in July 2026 that the checks “will not even attempt to make an assessment of what each customer can afford to gamble” and that operators would not be required to request bank statements following an FRA.

Jurisdiction	Prevalence Measure	Rate / Finding	Key Enforcement Response
Great Britain (UKGC)	PGSI 8+ (GSGB 2025)	2.4% of adults	Tiered FRA framework, product-level slot harm data drives scope
France (ANJ)	Algorithmic (23 indicators, H2 2025)	8.7% of online account-holders, 60% of online GGR	€500,000 fine (Decision 2026-031); algorithm as detection standard
Sweden (Spelinspektionen)	Channelisation / Spelpaus data (2025)	84% channelisation, 134,500+ Spelpaus registrations	Credit ban (1 May 2026); SIFS 2026:3 API mandate (1 Aug 2026)
Maryland (US)	State prevalence survey (2024)	~6% of residents (40%+ rise since 2022 legalisation)	Proposed credit card ban, AI-marketing restrictions under review
Pennsylvania (US)	Joint State Government Commission report (2026)	~25% of adults at risk of gambling disorder	Mandatory loss, deposit, and session-frequency limits proposed

How Does Prevalence Data Actually Drive Enforcement?

Prevalence surveys and algorithmic detection tools do not directly produce fines. The enforcement pipeline runs through a more specific mechanism: regulators use population-level data to establish what a competent operator should be detecting at account level, then audit operators against that standard. The gap between what the regulator’s model predicts and what the operator’s systems report becomes the evidentiary core of a compliance investigation.

The ANJ’s Company X decision is the clearest live example of this mechanism. The operator’s systems classified players at a lower risk tier than the ANJ’s own scoring model would have assigned. The regulator did not need to prove that those players were objectively problem gamblers under a clinical definition. It needed to show that the operator’s detection methodology failed to flag accounts that a reasonable compliance system, applying the 2021 ANJ reference framework, should have flagged. The fine is anchored to the detection failure, not to a retrospective finding of harm to specific individuals.

The UK’s Paddy Power Betfair enforcement action, which resulted in a £2 million settlement published in December 2025, demonstrates the same structure operating under the UKGC’s LCCP social responsibility codes. The Gambling Commission found that one customer deposited £12,000 in 15 days before being identified for review, another deposited £25,000 in 25 days, a third staked £86,000 over 16 days without any manual account review. The Commission’s casework note was explicit: the operator’s systems were “not sensitive enough to identify indicators of harm.” The prevalence of such cases across the UKGC’s enforcement caseload, not any single survey figure, is what drove the consultation on more prescriptive customer interaction thresholds.

Operators should understand this mechanism as a ratchet: each published prevalence finding or algorithmic estimate raises the floor for what regulators consider adequate detection. Operators calibrating their risk-scoring against their own historical enforcement thresholds, rather than against the regulator’s current published data, will find their systems outdated at the next inspection cycle.

Spelinspektionen’s Structural Response: From Data to Binding Obligations

Sweden’s response to its own prevalence and channelisation data illustrates how a regulator translates population-level findings into prescriptive compliance instruments. Spelinspektionen’s 2025 channelisation report, published on 15 June 2025, confirmed that the licensed market captured 84% of online gambling spend, down from 85% in 2024 and 86% in 2023. The report identified self-excluded players as the primary cohort using unlicensed platforms: individuals registered with Spelpaus who, once excluded from licensed sites, migrated to offshore operators that are not party to the self-exclusion system.

This finding produced two enforcement instruments. From 1 May 2026, all Swedish gambling licensees must ensure that customer deposits cannot be traced to credit cards, overdrafts, financial loans, or buy-now-pay-later services, a comprehensive credit ban under the Gambling Act 2018 (Spellagen 2018:1138). The obligation extends to external payment providers such as e-wallets: where an e-wallet itself extends credit or allows deferred payment, the licensee must block that payment route. The ban is the first of its kind across European licensed gambling jurisdictions.

From 1 August 2026, SIFS 2026:3 introduces mandatory API-based Spelpaus verification with unique Actor IDs and API Keys issued to each licensee. Verification is required before any direct marketing communication is sent, at new player registration, and at each player login. Marketing checks must use a dedicated marketing API, registration and login checks must use a separate login API. Spelpaus registrations now exceed 134,500 individuals, and the inspectorate’s move to real-time API verification converts self-exclusion from a passive safeguard into an active, auditable compliance duty.

The Methodology Dispute and Why It Does Not Reduce Compliance Risk

A recurring industry argument is that population-level gambling surveys over-estimate problem gambling prevalence because of sampling bias, topic salience effects, and the difficulty of distinguishing harmful from non-harmful high-frequency gambling. This argument has been made explicitly in response to the GSGB, and versions of it have circulated in relation to ANJ’s algorithmic estimates. Compliance officers should understand the argument’s limits.

Methodological criticism of a survey does not reduce an operator’s compliance obligations, because those obligations are grounded in the regulatory text: the LCCP social responsibility codes in Great Britain, the 2021 ANJ reference framework in France, and Spellagen together with Spelinspektionen’s regulations in Sweden. The survey data functions as context for regulatory supervisory priorities and as a benchmark for audit design. It does not, in itself, create the obligation to interact with at-risk customers, that obligation already exists in the primary regulation.

What the data does affect is enforcement intensity and targeting. Regulators that can demonstrate a credible population-level estimate of problem gambling are better positioned to justify intensive audit programmes, to allocate inspection resources toward high-risk product verticals, and to rebut operator arguments that their current detection rates are reasonable. The ANJ’s decision to publish its algorithm output before finalising the Company X enforcement action reflects precisely this sequencing: establish the benchmark publicly, then enforce against it.

The North American Dimension: Prevalence Data as Legislative Catalyst

The enforcement consequences of prevalence data are not limited to European licensed markets. A 2024 statewide survey in Maryland found that approximately 6% of residents reported gambling problems, an increase of more than 40% since the state legalised mobile and online sports betting in 2022. That finding is now cited in ongoing legislative debates about restricting credit card use for online betting and limiting AI-driven personalised marketing.

A 2026 report by Pennsylvania’s Joint State Government Commission found that an estimated one in four adults was at risk of developing a gambling disorder, set against a backdrop of record gaming revenues of $6.8 billion in 2025. The report recommended mandatory player limits on losses, play duration, and deposit frequency, along with restrictions on gambling advertisements near college campuses. New York launched a landmark 10-year longitudinal gambling behaviour study in 2026 under the State Office of Addiction Services and Supports (OASAS), collecting data via surveys, interviews, and focus groups across the state’s adult population. Regulatory interventions arising from its interim findings should be anticipated well before the study’s natural conclusion.

In Ontario, the Alcohol and Gaming Commission of Ontario (AGCO) fined theScore $105,000 under the Registrar’s Standards for Internet Gaming for failing to adequately monitor and intervene with a player who wagered $2.5 million over eight months while displaying clear signs of loss-chasing. The AGCO’s finding, that the operator relied on player self-assessments rather than conducting meaningful behavioural monitoring, mirrors the ANJ’s approach in Decision 2026-031 and the UKGC’s rationale in the Paddy Power Betfair settlement: the failure to detect is itself the violation, independently of whether the player’s harm can be individually quantified.

What Compliance Programmes Need to Recalibrate

The convergence across jurisdictions is not coincidental. Regulators are explicitly benchmarking their detection standards against published prevalence data, and they are designing enforcement cases around the gap between operator-detected populations and regulator-estimated populations. The operative compliance question is no longer whether an operator has a risk-scoring system in place, it is whether that system’s output is plausibly proportionate to the problem-gambling population in the operator’s player base.

Operators across France, Great Britain, and Sweden, and increasingly in North American licensed markets, should audit their risk-scoring systems against the following reference points. The ANJ’s 23-indicator framework, applied to H2 2025 data, classifies approximately 8.7% of French online account-holders as high-risk or excessive. If an operator’s own system identifies materially fewer than that proportion in a comparable player base, the operator faces an evidential deficit in any ANJ inspection. In Great Britain, the GSGB’s product-level harm data means that online slots players warrant higher-sensitivity thresholds within customer interaction frameworks under the UKGC’s LCCP social responsibility codes. In Sweden, any player whose Spelpaus status is not verified before each marketing communication, at registration, and at login is, from 1 August 2026, a SIFS 2026:3 compliance failure regardless of whether that player has displayed any other risk signals.

Operators should also scrutinise the internal separation between risk identification and intervention. ANJ Decision 2026-031 established that these are independently enforceable obligations: identifying a player as high-risk but failing to provide proportionate, graduated support is a separate violation from failing to identify them at all. Compliance programmes that treat risk-scoring output as complete, rather than as the trigger for a mandated intervention workflow, are structurally exposed in any jurisdiction applying the ANJ framework or its equivalents.

For a detailed examination of responsible gambling self-exclusion systems, deposit-limit architectures, and customer interaction obligations across all major regulated jurisdictions, the Responsible Gambling Compliance hub provides a cross-jurisdictional reference by topic stream. The SIFS 2026:3 technical requirements article sets out the Spelpaus API implementation obligations in detail for licensees approaching the August 2026 deadline. Operators with exposure across multiple European jurisdictions should consult qualified legal counsel on the application of jurisdiction-specific detection standards and intervention frameworks to their player bases.

Key Resources

Gambling Survey for Great Britain (GSGB), Wave 4, September 2025 to January 2026: Official statistics on gambling participation, published 26 February 2026 by the Gambling Commission. Available at gamblingcommission.gov.uk.

ANJ, Décision n°2026-031 du 10 juillet 2026: Full enforcement decision imposing a €500,000 fine on Company X for failing to identify and support high-risk players. Available at anj.fr.

ANJ, Algorithm publication (May 2026): “L’algorithme développé par l’ANJ révèle un nombre de joueurs excessifs inquiétant et en croissance.” Explains the 23-indicator methodology and H2 2025 findings. Available at anj.fr.

Spelinspektionen, SIFS 2026:3: Regulations on the national self-exclusion register (Spelinspektionens föreskrifter om det nationella självavstängningsregistret), decided 23 April 2026, effective 1 August 2026. Available at spelinspektionen.se.

UK Government, The Tax Treatment of Remote Gambling: Summary of Responses (2026): Cites Health Survey for England harm data and GSGB product-level findings on online slots. Available at gov.uk.