[Photo: Ricaela Diputado]
As artificial intelligence increasingly permeates the global gaming industry, a panel of experts at the G2E Asia + Asian IR Expo last week delivered a sharp reality check: AI can detect risk, but it cannot decide what is right.
The session, “From Vision to Action: How AI Is Reshaping Responsible Lottery and Gaming,” brought together researchers, operators, and technology developers at the Venetian Macao to confront a central challenge: the industry is improving at identifying problem gamblers, but meaningful intervention remains elusive.
“We’re entering a very interesting phase,” said Travis Sztainert, director of research and education at the International Center for Responsible Gaming (ICRG). “For the past several years, the focus has really been on detection. But the future of AI and responsible gambling is not just about identifying risks. It’s about what we do after we identify them – the so what and now what?”
Sztainert predicted that within one to two years, AI will shift from “primarily surveillance” to “supporting infrastructure,” including personalized interventions, AI-assisted messaging, and real-time synthesis of research. Operators and regulators, he said, will stop asking, “Can AI do this?” and start asking, “How should AI do this responsibly?”
Two shifts ahead
Wayne Gao, chief AI officer of GENLOT, outlined two major shifts. First, intervention will move from reactive to anticipatory.
“Today, players need to cross a threshold for the system to act,” Gao said. “With AI agents, the system sees the drift early and could intervene slightly. The player experience changes from getting flagged to being supported.”
Second, he noted a divergence between lottery and casino operations. “The public welfare narrative becomes real-time,” he said. “Today, players know in abstract that lottery funds are a public good. With AI, each transaction could carry that meaning – not a slogan anymore.”
Gao warned that “AI always gets sold as a magic solution. It is not. AI changes what’s possible. But humans still decide what is right.”
The “black box” problem
Yardena Almagor, customer success lead at Mindway AI, emphasized early detection of emerging risks rather than focusing only on high-risk players. She said personalized messaging is essential.
“AI needs to explain what the problem is, where the issues are, what we’ve identified in the customer’s behavior,” Almagor said. “It’s not just a black box – here it is, high risk, medium risk, and that’s it.”
Michael Auer, managing director of Neccton of OpenBet Limited, questioned the industry’s fundamental goals.
He said operators fear revenue loss. “We as researchers should ask ourselves how we can help operators make more money from healthy players.”
Gao identified three barriers to deploying AI agents in gaming: data fragmentation across system vendors and regulatory functions; an auditability gap in which current AI models do not produce traceable decisions; and trust transfer from human operators to AI.
“Floor staff have spent decades developing instinct,” Gao said. “We can’t just put an AI in front of them and expect them to defer. We need to build a system gradually to earn trust –first show the reasoning, allow humans to overrule it, then learn from the overrules. This is an organizational problem, not a technical problem.”
Standards challenge
Professor Guojing Su, founder of the China Lottery Industry Salon and president of the Macao Association for the Promotion of Moderate Diversification of Economy, called for urgent standardization.
“The most important thing is a standard. Every step of AI-driven data digestion – what are the standards for the entire system structure?” Su said. “Macau should build a standard system that is suitable for localization as soon as possible.”
He also pointed to emerging real-world complications, citing AI glasses that became available last year.
“Can these AI glasses be brought into the venue? Theoretically, it is not allowed. But what is the standard? Can I really wear glasses? It involves privacy and many other aspects. Policies need real-time adjustment.”
Sztainert called for integrating responsible gaming ecosystems. “Right now, many systems operate separately – risk detection, messaging, compliance, customer support. They’re not working together. Connect them.”
Almagor said data remains a key challenge, particularly in land-based gaming.
“Our models are only as good as the data that comes in,” she said.
Gao agreed, saying data collection and interpretation must be tailored for AI-driven systems.
Auer made an open offer, saying he wants to align player protection with commercial goals.
“I can help you earn more money with fewer problematic players,” he said.
Su said that once a standard system is in place, experts can help develop practical solutions.
Ryan Ho, lecturer and program coordinator of the Center for Gaming and Tourism Studies at Macao Polytechnic University, who moderated the panel, concluded: “We must be building a world that is safe, smart, fair, and fun for everyone.”
No Comments