Land-based casinos test AI for safer gambling tools: Expert
Land-based casinos are increasingly exploring AI-based player protection tools, according to an expert. But industry data suggests technology adoption remains uneven, governance is still developing, and many operators are still working out how to translate automated risk scores into effective intervention strategies.
Speaking at G2E Asia in Macau, Yardena Almagor, Customer Success Lead at Mindway AI, outlined how behavioural AI models are being used to move beyond traditional threshold-based monitoring, while acknowledging the structural limitations of current industry approaches.
“When people hear AI, they really think it’s complicated and something very fancy and a little bit even magic,” she said. “But in reality, it’s much simpler than that.”
Her comments come as the broader gambling sector is still in the early stages of AI maturity. According to the State of AI in Gaming 2026 benchmark report, developed by the University of Nevada, Las Vegas International Gaming Institute in collaboration with KPMG LLP, the industry’s average AI maturity score stands at 45 out of 100, indicating early-stage adoption and ongoing governance gaps.
The UNLV report was based on responses from 83 gambling companies and 113 regulators worldwide. It found that while most operators have strategic ambitions around AI, many lack the infrastructure and expertise needed to scale implementation effectively. It also noted that land-based operators lag behind online counterparts.
According to the report, AI use is concentrated in technology, security, and product innovation, which account for nearly half of all initiatives. While generative AI is widely used, over 80 per cent of companies apply it for content and insights, more advanced “agentic” AI remains limited due to compliance and player safety concerns.
Threshold systems under pressure
Almagor highlighted a key challenge shared by many operators, the reliance on thresholds such as spend limits, self-exclusion triggers and behavioural spikes. She said that while these are widely used, they do not capture the full spectrum of gambling harm.
“We know that only 20 per cent of people who are problem gamblers actually self-exclude,” she said. “And we also know that only 30 per cent of those that actually self-exclude are problem gamblers.”
“Problem gambling doesn’t just appear from one action,” she said. “It’s a relation over time of lots of different changes in behaviour, sometimes very, very nuanced.”
The UNLV IGI report notes that many organisations struggle to define success in AI-driven systems, with only 1 in 5 reporting a meaningful return on investment within 2 years. One in four companies also lacks a structured evaluation process for AI tools, making performance difficult to measure.
Land-based data highlights behavioural complexity
Almagor referenced an analysis conducted with an Australian land-based casino operator, in which behavioural models were applied to nearly 200,000 players over a four-month period. The dataset included session activity, player profiles and self-exclusion records.
The study found that spending alone offered limited insight into risk. Two high-value “platinum” customers, for example, displayed similar monetary behaviour but very different gambling patterns.
“One player exhibited erratic and intense play across multiple games,” she said, while another showed more consistent and stable behaviour.
Governance, responsibility gaps remain
Despite growing interest in AI-driven monitoring, UNLV’s State of AI in Gaming 2026 report highlights major gaps in the responsible implementation of AI across the sector.
Nearly one-third of gambling companies report having no formal responsible AI framework, while only around 2 per cent have fully embedded responsible AI practices across their organisations. A further third describes their efforts as “starting” or “basic.”
The report also found that fewer than 20 per cent of organisations have dedicated AI governance roles, with oversight often fragmented or shared across multiple departments. In some cases, no formal accountability structure exists at all.
Regulatory confidence is also limited. The report notes that regulators often lack visibility into how licensees deploy AI, creating a disconnect between perceived and actual use cases. While regulators tend to focus on customer-facing applications, most AI activity is concentrated in back-end technology, security and product systems.
Human oversight is still central to intervention
Almagor stressed that AI systems should support, not replace, human decision-making in responsible gambling frameworks. A low-risk player may receive general safer gambling messaging, while higher-risk players may trigger direct staff engagement.
“That high-risk player needs someone to reach out to them physically and go and speak to them,” she said.
Shape your feed around the future of gaming. Choose SiGMA World as your preferred source for insider voices, prediction market coverage, casino developments and global regulation news.
