AI in online gambling: regulators draw lines between engagement and protection

The rules around artificial intelligence in online gambling took clearer shape on several fronts between 2025 and 2026, even though no regulator has banned personalisation outright. The EU AI Act's bans on harmful manipulation have applied since 2 February 2025, the Dutch regulator has said it will issue guidelines on AI and automated monitoring in 2026, and in Great Britain operators have been required since 2022 to use automated processes to act on signs of harm. The same technology that tailors a lobby is increasingly expected to protect the player looking at it.

This piece sets out what operators use AI for, which rules now apply, what research says about harm-detection models, and what is still unsettled. It relies on regulator texts, one expert-interview study, one peer-reviewed prediction study and the published pages of named companies.

One technology, two jobs

In online gambling, AI is used for two broad purposes. The first is engagement: ranking games in a lobby, choosing which offer to show and when, and running support chat. The second is protection: spotting patterns linked to harm, such as rising deposits or chasing losses, and prompting an intervention. The data behind both is largely the same account history.

A study published in the Journal of Gambling Studies (volume 42, March 2026) by Binesh, Ponnada and Syah interviewed 41 industry professionals in two rounds, in 2021 and 2023. The authors describe a shift from experimental use of AI towards personalised interactions and predictive analytics, and list regulatory gaps, algorithmic bias and responsible gambling concerns among the main challenges.

How the personalisation loop works, and where UK rules cut into it

Personalisation is a feedback loop: behaviour feeds a model, the model changes what the player sees, and the player's response becomes new data. In Great Britain, two sets of rules act directly on the offer side of that loop. From 1 May 2025, online operators must let customers opt in to direct marketing by product (betting, casino, bingo) and by channel, with opt-out as the default.

From 19 January 2026, bonuses may carry wagering requirements of no more than 10 times, and promotions that mix products, such as betting and slots in one offer, are banned. However sophisticated the targeting, the offer it can deliver is now narrower.

Great Britain already requires automated harm detection

On the protection side, the UK rules go further than many players realise. Under social responsibility code 3.4.3, which applied in stages from 12 September 2022, remote operators must flag indicators of harm in a timely manner and act on strong indicators by implementing automated processes. Where strong indicators are identified, they must prevent marketing and the take-up of new bonus offers.

The same code builds human oversight into any automated step: operators must manually review how the process worked in each customer's case and give customers "the opportunity to contest any automated decision which affects them".

On 5 January 2026 the Commission published its own approach to AI. It names understanding and, where appropriate, regulating the industry's use of AI as a priority, and says its own use will be lawful, transparent and subject to human oversight. It has not published AI-specific rules for operators.

The Netherlands: chatbots allowed, humans required

The Dutch regulator, the Kansspelautoriteit (KSA), wrote to licensed online operators in May 2025 about chatbots and AI. As reported by the trade site CasinoNieuws, the letter said AI may support routine customer service but cannot replace trained staff in addiction prevention, and that operators remain responsible for systems they buy in and must be able to explain how they reach conclusions.

KSA chair Michel Groothuizen said on 16 January 2026 that the regulator would issue guidelines on the use of AI and automated monitoring tools later in the year, and the KSA's 2026 supervision agenda lists that guidance under its duty-of-care work.

The EU AI Act timetable after the July 2026 omnibus

The EU AI Act entered into force on 1 August 2024 and applies in stages. Its prohibitions, which include harmful AI-based manipulation and harmful exploitation of vulnerabilities, have applied since 2 February 2025. The AI Omnibus, which entered into force on 27 July 2026, moved the start of rules for stand-alone high-risk systems to 2 December 2027.

The Act does not name gambling. Whether a particular casino recommender or bonus engine could fall under the manipulation or vulnerability bans is a legal question that, as far as we could find, no regulator or court has yet tested.

Does harm-detection AI work?

The evidence is promising but limited. In a study in the Journal of Gambling Studies, Michael Auer and Mark Griffiths trained models on the account data of 945 players at one European online casino to predict self-reported problem gambling. The best model, a random forest, reached an AUC of 0.729 on test data. The authors list clear limits: one operator, a short observation window, and self-reported answers.

Operators and suppliers publish less detail. Entain describes its ARC programme as using AI to learn from and identify risks in player behaviour, and Mindway AI said in March 2023 that its GameScanner tool is part of ARC. SOFTSWISS said in October 2025 that it uses a machine-learning risk scoring tool and handled 16,000 responsible gambling cases in the first half of 2025. None of these companies has published independent evaluations on those pages.

What is confirmed and what is not yet known

• Confirmed: UK operators must use automated processes on strong indicators of harm, with manual review and a right to contest; AI Act prohibitions have applied since 2 February 2025; the KSA expects human staff in addiction prevention.
• Not yet known: the content of the KSA's promised AI guidelines, which we had not found published at the time of writing; whether any gambling system will be treated as a prohibited or high-risk practice under the AI Act.
• Unclear: how widely operators use AI for personalisation and how well their harm models perform; published figures come mostly from the companies themselves.

What it means for players

Personalisation changes what you are shown, not how a game pays: a slot's return to player is set by its design, as our RTP and volatility guide explains. In Great Britain you can choose which products and channels you get marketing for, and bonus terms are capped. If an operator contacts you about your play, it may have come from an automated flag, and UK rules entitle you to a human review. Deposit limits, time-outs and self-exclusion remain the controls you set yourself; see our responsible gambling guide.

Frequently asked questions

Sources

1. European Commission - AI Act regulatory framework
2. European Commission - AI Omnibus enters into force (27 Jul 2026)
3. UK Gambling Commission - The Commission's approach to Artificial Intelligence (5 Jan 2026)
4. UK Gambling Commission - LCCP 3.4.3 Remote customer interaction
5. UK Gambling Commission - New rules on action for at-risk customers (14 Apr 2022)
6. UK Gambling Commission - New rules boosting safety and consumer choice (1 May 2024)
7. UK Gambling Commission - LCCP 5.1.12 Direct marketing preferences
8. UK Gambling Commission - Gambling promotions to be safer and simpler (26 Mar 2025)
9. Kansspelautoriteit - Speech Michel Groothuizen (16 Jan 2026)
10. Kansspelautoriteit - Toezichtagenda 2026
11. CasinoNieuws - KSA letter on chatbots and AI (13 May 2025)
12. Binesh, Ponnada and Syah - The Future of the Gambling Industry is AI, Journal of Gambling Studies (Mar 2026)
13. Auer and Griffiths - Predicting self-reported problem gambling with account data, Journal of Gambling Studies
14. Entain - Protecting players with ARC (2021)
15. Mindway AI - GameScanner and Entain ARC (15 Mar 2023)
16. Yogonet - SOFTSWISS machine learning for responsible gambling (8 Oct 2025)