KSA launches open-source model to assess risky online gambling

The Netherlands Gambling Authority, known as the Kansspelautoriteit (KSA), has highlighted a new open-source machine-learning model developed by researchers at the University of Amsterdam (UvA) to help assess risky gambling behaviour among online casino players. The initiative is focused on analysing actual playing patterns and is intended to provide operators with an additional instrument for meeting their duty-of-care responsibilities.

The KSA reported the development on 18 August 2026 as part of its continuing work on player protection and the supervision of online gambling. The regulator's wider approach places significant importance on monitoring player behaviour and ensuring that licensed operators respond appropriately when indications of risky or potentially problematic gambling emerge.

Model analyses actual player behaviour

The new model is designed to assess risk by examining behavioural data generated through online gambling activity. Rather than relying primarily on what a player says about their own habits the approach looks at observable patterns in how an account is used.

The model considers factors such as betting behaviour, gambling frequency and the timing of playing sessions. It also examines sequences of wins and losses and the way a player's activity changes around those periods. These signals are processed by the model to estimate a level of potential risk.

This behavioural approach is significant because online gambling platforms generate large volumes of activity data. Playing frequency, session timing, betting intensity and changes in behaviour can provide a more continuous picture of activity than isolated questionnaires or occasional assessments.

The model is intended to help operators identify situations that may warrant additional attention. A risk assessment does not by itself establish that a player has a gambling disorder or that an intervention is necessarily required. Instead, it can provide another data point for operators when they assess the wider circumstances of an individual account.

The project covers online gambling activity broadly rather than being restricted to a single game category. That makes the model potentially relevant across different forms of remote gambling where behavioural information can be collected and assessed consistently.

KSA presents the model as a supporting tool

A central element of the initiative is the distinction between technological assistance and regulatory responsibility. The KSA has not presented the model as an autonomous system that can replace an operator's own assessment or duty-of-care procedures.

This distinction is particularly important in the Dutch regulatory environment. The KSA has previously emphasised that online gambling providers must monitor playing behaviour and act when relevant indicators of problematic participation are identified. Its supervisory work has also examined the speed and quality of behavioural analysis and the actions taken after risk signals appear.

The machine-learning model can therefore be understood as an additional analytical resource. Operators may use it to support their own monitoring frameworks and compare its results with other signals available through their player-protection systems.

The KSA's latest supervisory agenda also indicates that the regulator is paying close attention to artificial intelligence and monitoring tools used by gambling operators. This reflects a broader shift toward more data-driven oversight in which technological systems are increasingly used to identify patterns that may not be obvious through manual review alone.

Open-source design could support wider use

The open-source element of the project could be particularly relevant for the wider gambling sector. By making a model available for examination and potential use researchers and operators can evaluate how the system works and consider how it may fit within existing monitoring processes.

Open-source availability can also support greater scrutiny of analytical methods. In a regulatory environment where automated systems may influence decisions affecting players transparency around methodology can be important. Operators still need to determine how an external model should be integrated into their own systems and governance processes.

The availability of a model does not remove the need for professional judgement. Behavioural indicators can have different explanations and a particular pattern may not necessarily point to harmful gambling. A responsible assessment therefore requires context and appropriate follow-up rather than reliance on a single numerical output.

This is also important from a legal and compliance perspective. Automated risk scoring can support decision-making but it should be used within applicable regulatory requirements and appropriate data-governance frameworks. Operators remain responsible for the manner in which player information is processed and how intervention decisions are made.

Dutch approach reflects broader regulatory developments

The KSA initiative forms part of a wider European move toward data-led approaches to player protection. Regulators in several jurisdictions are exploring ways of using behavioural information to identify potential risk earlier and to create more consistent benchmarks for operator monitoring.

Spain's Directorate General for the Regulation of Gambling (DGOJ) has also developed a mechanism designed to detect risk-related behaviour in online gambling. Its system uses supervised machine learning and real-world behavioural data to identify patterns associated with risky play. The Spanish regulator has described the mechanism as part of its broader player-protection framework.

France has taken another significant step. In May 2026, the Autorité Nationale des Jeux (ANJ) announced results from its algorithm for estimating excessive gambling in online and account-based gambling. The regulator said the model identified around 600,000 players with a high probability of excessive gambling during the second half of 2025. That represented 8.7% of the registered player population and those players accounted for approximately €1.2 billion in GGR, equivalent to 60% of total GGR under the methodology specified by the ANJ.

The ANJ has made its algorithm available as an optional tool that operators can use alongside their own systems. It also views the model as a benchmark for evaluating how effectively operators identify excessive gambling.

These developments suggest that regulators are increasingly interested in approaches that move beyond general responsible-gambling messaging toward structured assessments based on observed account behaviour.

Technology does not remove operator responsibility

For operators the introduction of behavioural models may bring both opportunities and additional compliance considerations. More advanced analytics can potentially improve the identification of unusual or escalating patterns but any intervention still needs to be proportionate and grounded in the applicable regulatory framework.

The quality of a risk assessment can depend on the data available to the model, the way signals are interpreted and the circumstances in which the resulting score is used. A model should therefore be considered part of a wider player-protection framework rather than a substitute for human review.

The KSA's own research into duty-of-care practices has highlighted the importance of monitoring multiple behavioural indicators rather than relying on a single measure. The regulator has also stressed the need for operators to respond appropriately when risk becomes apparent.

That context gives the new model a practical role. Its purpose is not simply to produce a score but to help create better visibility of behaviour that may otherwise require more time or resources to identify.

Implications for future gambling regulation

The development of the KSA model could contribute to a more standardised discussion about how behavioural risk should be assessed across online gambling markets. Regulators can use common analytical frameworks to compare trends while operators can assess whether their internal systems are identifying risk consistently.

At the same time the growing use of machine learning will likely increase scrutiny of transparency, data quality and accountability. Questions around how models are validated, how false positives are handled and how operators document decisions may become increasingly important as automated tools become more common.

For the Dutch market the initiative also reinforces the KSA's broader objective of strengthening player protection through earlier recognition of potentially risky behaviour. The regulator has made duty-of-care supervision a continuing priority and has indicated that it will continue assessing how online operators analyse playing activity and intervene when risks are detected.

Conclusion

The KSA's support for an open-source machine-learning model marks another development in the use of behavioural analytics for online gambling oversight. Created by University of Amsterdam researchers and focused on real playing behaviour the model offers operators another way to identify patterns that may warrant closer attention.

Its significance lies less in automation alone than in the potential to strengthen the broader duty-of-care process. A risk score cannot determine a player's circumstances by itself and it does not remove the responsibility of licensed operators to assess behaviour carefully and take appropriate action.

The Dutch initiative also places the KSA within a wider European movement toward evidence-based player protection. Similar work by the DGOJ in Spain and the ANJ in France shows that regulators are increasingly exploring analytical tools that can provide common benchmarks and improve the visibility of risky gambling patterns.

As these systems develop the central regulatory principle is likely to remain unchanged: technology can support responsible gambling oversight but accountability ultimately remains with the operators and regulators responsible for protecting players.

FAQs

What is the new KSA model?
The KSA model is an open-source machine-learning tool developed by University of Amsterdam researchers to assess potential risky gambling behaviour using actual online gambling activity.

What information does the model analyse?
The model examines behavioural signals such as betting patterns, gambling frequency, session timing and reactions to winning and losing streaks.

Is the KSA model mandatory for online gambling operators?
The model is presented as a supporting resource rather than a replacement for operators' existing duty-of-care systems. Operators remain responsible for complying with applicable requirements.

Can the model determine whether a player has a gambling disorder?
No. A behavioural risk score does not by itself establish that a player has a gambling disorder. It is intended to provide an additional indicator for further assessment.

Why is open-source availability important?
An open-source approach can allow operators and researchers to examine the model and consider how it can be incorporated into broader monitoring and player-protection frameworks.

Does the model replace human assessment?
No. Human judgement and wider contextual assessment remain important when determining whether a player may require an intervention.

Is the project limited to online casinos?
The project is intended to cover online gambling broadly rather than being confined to one particular game type.

Are other European regulators developing similar tools?
Yes. Spain's DGOJ and France's ANJ have also developed or introduced behavioural and algorithmic approaches for identifying gambling-related risks.

What has the ANJ reported about its algorithm?
The ANJ said its algorithm identified approximately 600,000 registered players as having a high probability of excessive gambling in the second half of 2025. It said those players represented 8.7% of registered players and generated approximately €1.2 billion in GGR under its stated methodology.

Why is behavioural monitoring important in online gambling?
Online platforms can generate detailed information about playing activity. Analysing those patterns can help operators identify changes or combinations of signals that may justify closer review and potential player-protection measures.