Prediction markets may be experiencing rapid growth, but for Jason Fama, SVP of Engineering at Sift, the industry’s long-term commercial prospects will depend on more than attracting new participants and expanding into new markets.
As concerns around insider trading, market manipulation and fraudulent activity intensify, Fama believes operators face a growing challenge: ensuring the integrity of their platforms can keep pace with their commercial ambitions.
Fama warns that the commercial consequences of inadequate fraud prevention could emerge long before regulators intervene.
He explained: “Sooner than most people think. It isn’t when a regulator steps in. It’s when outsiders can measure the gap between what a platform says about itself and how its markets actually behave.”
Speaking to SBC Americas, Fama examines how the unique characteristics of prediction markets are creating new opportunities for fraud, why traditional sportsbook controls may not be sufficient and how AI-powered detection could help operators turn market integrity into a competitive advantage.
When integrity becomes a commercial liability
For prediction market operators, the financial consequences of fraudulent activity extend well beyond the immediate losses associated with individual incidents.
According to Fama, the greater threat lies in the gradual erosion of confidence among customers, commercial partners and regulators.
He said: “Direct losses are honestly the smallest part. The bigger damage is cumulative.
“Everyday users start to feel they’re trading against people who know more than they do, and that perception travels faster than any marketing correction. They leave, or they never sign up.”
Recent controversies have demonstrated how quickly questions surrounding market integrity can attract wider scrutiny.
Fama pointed to reporting by 60 Minutes concerning nine connected Polymarket accounts that reportedly achieved a 98% win rate on military-related contracts, alongside allegations involving the use of classified information to trade on prediction markets.
Such incidents raise questions about whether platforms can identify participants operating with privileged information or exploiting weaknesses in market design.
However, Fama suggested that the repercussions extend beyond consumer confidence.
He explained: “Partners, whether media companies, payment providers, or distributors, get nervous about attaching their brands to a market with open integrity questions.”
Why prediction markets present different fraud risks
Although prediction markets share certain characteristics with traditional sports betting, Fama believes their underlying structures introduce a broader range of potential vulnerabilities.
Sportsbooks have established systems to address payment fraud, promotional abuse and account takeovers. Prediction markets, however, must also consider whether participants can influence the outcomes of the events themselves.
“The biggest difference is that the thing being manipulated may not be the price,” Fama explained. “It could be the event itself, the evidence used to settle it, or who holds an information advantage.”
He highlighted an incident involving Spotify‘s music charts, where more than 500,000 artificial streams reportedly distorted the results used by Kalshi to settle a $3m prediction market contract.
Fama added: “The contract types are broader, the settlement mechanisms vary more, and the attack surface is harder to define.”
Can AI detect fraud without damaging customer experience?
As prediction market platforms expand, identifying suspicious activity without introducing unnecessary friction for legitimate participants represents another operational challenge.
For Fama, the solution lies in applying greater scrutiny to activity that demonstrates identifiable risk rather than subjecting every customer to additional verification.
He stated: “Put the scrutiny where the risk is. Most users should never notice the controls.
“The signals worth watching are clusters of connected accounts or wallets, unusually confident positions ahead of an event, concentrated profits, and strange win rates on long shots.”
He also advocated adjusting monitoring requirements according to the nature of individual contracts, arguing that markets involving military operations warrant greater scrutiny than routine economic announcements.
This risk-based approach is an area where AI-powered behavioural analysis could become important.
At Sift, Fama explained, the focus is on identifying relationships between signals that may not appear suspicious when examined individually.
He explained: “Shared devices, linked identities, and trading patterns that look too coordinated are hard to see account by account but much easier to see in aggregate. That’s the kind of problem we spend our time on at Sift.”
However, AI presents challenges alongside opportunities.
The same technology that enables businesses to identify coordinated behavior can also make synthetic identities and automated fraudulent activity easier to create at scale.
“It’s an arms race,” Fama warned. “The goal should be continuous detection that feeds back into contract design, access controls, and market approval, not just account enforcement.”
What prediction markets can learn from sportsbooks
For prediction market operators seeking to strengthen their integrity frameworks, the established US sports betting industry provides a potential starting point.
Fama acknowledged the investment sportsbooks have made in identity verification, geolocation, responsible gaming and collusion monitoring, suggesting prediction markets should adopt relevant elements of these existing practices.
“There’s no reason to reinvent that wheel,” he stated.
Nevertheless, he cautioned that traditional sportsbook safeguards cannot address every vulnerability associated with prediction markets.
Whereas conventional sporting outcomes are generally determined independently of bettors, prediction market participants may possess privileged information, influence an underlying event or attempt to manipulate the evidence used to settle a contract.
Measuring the commercial value of prevention
As operators weigh the costs of strengthening fraud prevention, Fama argued that measuring success purely through prevented financial losses risks overlooking its wider commercial contribution.
Instead, he recommended evaluating detection accuracy, operational efficiency and the impact of security measures on legitimate customers.
He explained: “I’d start with detection quality: false positive rates, time to detect, and how much gets caught before settlement rather than after.”
Fama also suggested monitoring market-specific indicators, including profit concentration, synthetic activity and anomalous win rates, alongside customer retention.
He stressed: “If the other numbers look good but casual users are quietly leaving, something is wrong.”
Ultimately, he believes the ability to demonstrate credible market integrity could distinguish successful prediction market providers as the sector matures.
Fama concluded: “Absolutely, and I think it’s the most underappreciated opportunity in the category. Participants assume markets are monitored and that serious abuse has consequences. Operators have to be able to show that’s true.”
