Singapore’s Gambling Regulatory Authority is making artificial intelligence part of its regulatory operating model. The regulator’s newly released 2025/26 annual report describes an AI Masterplan, wider use of digital tools and systems designed to identify anomalies in audit-log reviews.
The important point is not that a regulator has adopted Microsoft Copilot or deployed internal bots. It is that Singapore is placing automation inside a risk-based framework: routine and lower-risk work can move faster, while human regulatory attention is concentrated where the consequences are higher.
What the GRA has disclosed
According to the 2025/26 report, GRA has expanded the use of Microsoft Copilot and other digital tools, introduced bots that help officers access internal resources such as finance regulations, and used technology to flag anomalies in audit logs. Chief Executive Daniel Tan said the authority’s AI Masterplan will guide how data and AI support regulatory functions, organisational effectiveness and staff capability.
The figures and initiatives were detailed by Inside Asian Gaming on 8 October 2026. The regulator publishes its reports through the official GRA annual reports archive.
This is an extension of an existing direction rather than a sudden AI pivot. In its previous annual report, GRA said its Digitalisation Masterplan 2.0 would integrate generative and analytical AI into applications, strengthen risk-based decisions and automate parts of compliance monitoring. The authority has also previously described using data and AI to identify characteristics associated with higher vulnerability to gambling harm.
Risk-based regulation is the real operating change
Singapore’s approach matters because regulatory capacity is finite. A regulator that treats every application, log entry and control test as equally risky can become slow without necessarily becoming more effective.
GRA’s report gives a concrete signal of the alternative. The authority says 40 projects implemented during FY2025 saved more than 2,600 staff hours. It also introduced a whitelist of approved casino marketing phrases and a streamlined route for lower-risk advertising and promotion applications, according to reporting on the annual report.
That does not mean oversight is being relaxed. It means the regulator is trying to reserve judgement-heavy work for the areas where it adds the most value. Singapore’s latest Casino Control amendments, which took effect on 21 September 2026, expanded regulatory powers and introduced additional offences. Technology is being added alongside a broader legal toolkit, not in place of it.
What operators and suppliers should expect
For licensed businesses, AI-assisted oversight changes the economics of weak controls. Large volumes of logs and repetitive records have historically made some anomalies expensive to find. Better triage can make inconsistent access, unusual system activity or recurring control failures more visible.
That should raise the value of clean data lineage and explainable systems. An operator may be able to show that a control exists; it will increasingly need to show how the control produced a decision, which data fed it and who reviewed the exception.
Suppliers should also expect more precise questions during approvals and audits. If regulators can search and compare evidence more efficiently, generic policy documents become less persuasive. The differentiator becomes whether the operational record matches the written control.
Our view: AI will accelerate scrutiny before it automates judgement
Our view is that the near-term impact will be faster regulatory triage, not automated enforcement. GRA’s disclosed use cases focus on finding information, reviewing logs and identifying anomalies. Those are support functions that help an officer decide where to look. They are not evidence that an algorithm is making licensing or enforcement decisions.
That distinction is commercially important. Operators should not respond by building a separate “AI compliance” narrative. They should improve the quality, consistency and retrievability of the underlying evidence that any human or machine-assisted review will encounter.
We expect Singapore’s model to be watched by other ASEAN regulators because it offers a practical route between two weak extremes: manual oversight that cannot keep pace with digital businesses, and black-box automation that is difficult to challenge or explain.
The uncertainty lies in implementation. GRA has not published technical thresholds, model-governance detail or error rates for the anomaly tools. Nor should an annual report be read as proof that every deployment is mature. The next meaningful evidence will be how these systems change regulatory turnaround times, the specificity of findings and the expectations placed on licensees.
For now, the signal is clear: Singapore wants a regulator that is human-led but increasingly machine-assisted. Follow IGI’s Singapore coverage and regulation reporting for the operational consequences.
