Evenbeter, formerly EvenBet Gaming, has launched an updated back-office platform that adds AI-assisted screening of suspicious poker and casino activity, tournament validation, demand forecasting and reporting assistants. The company says the tools are designed to help operators decide what deserves attention, but the key question for fraud and compliance teams is whether automated flags can improve detection without becoming opaque or undermining fair review and data privacy. The launch materials describe safeguards such as human decision-making and restricted data scopes, yet no independent performance or audit evidence has been published.
How the new suspicious-activity screening works
According to Evenbeter, the AI-assisted screening can flag suspicious hands and account connections for detailed review. It analyses player or tournament hand histories from the bubble stage onwards alongside internal signals such as shared devices and payment links. The company says those signals are processed securely without personal data like names or IP addresses, and the system highlights potential multi-accounting and coordinated activity to support human investigators without establishing wrongdoing on its own.
The screening is one part of a broader update that includes tournament validation, demand forecasts with confidence ranges, natural language report filters and responsible gaming behavioural markers. AGBrief reported those features as an extension of the existing back office rather than a separate fraud engine.
Human review remains central, not optional
Alexander Tamplon, CTO at Evenbeter, said: "An operator needs to understand what deserves attention and what to check next. We are bringing that support directly into the back office: highlighting suspicious hands and account links, checking tournament settings before launch, and helping teams interpret reports. The tools provide context and recommendations, while operators remain responsible for the decisions."
Administrators retain full governance, with the ability to enable individual AI features, configure daily request limits, tailor assistant instructions, and strictly control data scopes according to back-office permissions. That is the most important safeguard for fair review: a generated flag is a lead to investigate, not a finding of misconduct. Evenbeter has been explicit that the screening does not automatically determine guilt. However, the company announcement does not describe independent oversight of these controls, so how fairly any deployment works will depend on an operator's configuration and investigative discipline.
Privacy claims are described, not externally verified
The screening processes internal signals such as shared devices and payment links without names or IPs, according to Evenbeter. That is a meaningful data-minimisation step, but the launch materials do not mention any assessment by data protection regulators or independent privacy audits, and they do not say whether operators will receive documentation to support their own compliance reviews. As a result, operators cannot assume the tool meets every market's privacy rules based on the announcement alone.
Soloazar reported the same data handling point, but the public announcement stops at the product description. The governance settings give administrators control over data scopes, which can help operators align the system with their own privacy obligations, but that control still depends on correct configuration and ongoing review.
Fraud detection gains are plausible, not proven
The update offers operational benefits that could indirectly support fraud detection: screening helps investigators prioritise suspicious hands and account connections, tournament validation can prevent schedule and payout errors, and demand forecasts can inform resource planning. But the launch materials provide no detection rates, false-positive rates, case study results or independent benchmarking. It is therefore not yet known whether AI-assisted screening will generate more accurate fraud detection than existing manual or rules-based methods.
That lack of evidence does not mean the tools will fail. It means operators and regulators should treat the launch claims as vendor descriptions, not as validated outcomes. Any operator considering the platform would need to test it against its own player base, fraud patterns and responsible gaming requirements before concluding that AI-assisted screening improves detection or reduces risk.
What would preserve fair review in practice
A fair deployment would need clear policies for human review, player notification where applicable, retention limits and audit trails. The update's administrator controls are a starting point, but operators will need to document why a flag was raised, what a human investigator found, and how decisions are communicated. If investigators treat flags as leads rather than presumptive evidence, AI-assisted screening can support a fair process.
The reader question is therefore best answered as a conditional: the design allows for privacy-conscious screening and human oversight, but no published evidence demonstrates that it improves fraud detection, and the fairness and privacy outcomes will depend on individual operator configuration, not on the product alone.
