Sift executive warns fraud risks threaten prediction market growth
Sift's Jason Fama tells SBC Americas that market integrity, not regulation, will determine whether prediction platforms keep users and partners.

Prediction markets face a commercial risk that may arrive well before any regulatory intervention, according to Jason Fama, SVP of Engineering at Sift, the fraud prevention firm.
Speaking to SBC Americas, Fama said the critical moment for operators comes not when a regulator acts, but when outsiders can measure the gap between how a platform presents itself and how its markets actually behave.
Fama argued that direct financial losses from fraud are the smallest part of the problem. The larger damage, he said, is cumulative: users who come to believe they are trading against better-informed participants will leave quietly, or never join at all, and that perception spreads faster than any attempt to correct it through marketing.
He pointed to reporting that found nine linked accounts on Polymarket, which reportedly won 98% of the time on contracts tied to military events, as well as allegations that classified information had been used to trade on prediction markets, as examples of the kind of scrutiny that can quickly damage confidence.
Fama said the fallout is not limited to individual users. Media companies, payment providers and distributors grow wary of attaching their brands to a market where integrity is in question, he told SBC Americas.
Prediction markets, Fama argued, carry a broader set of vulnerabilities than traditional sports betting. Sportsbooks have built systems to tackle payment fraud, promotional abuse and account takeovers, but prediction markets must also guard against participants influencing the outcome of the underlying event itself.
He said the object of manipulation in these markets may not be price at all, but the event, the evidence used to settle it, or an information advantage held by certain participants.
As an example, Fama noted that Kalshi had used Spotify chart data to settle a prediction market contract worth $3m, and that more than 500,000 fake streams reportedly skewed the figures relied on for that settlement. He said contract types in this sector are broader, settlement mechanisms vary more widely, and the attack surface is correspondingly harder to define.
On detection, Fama said the priority should be applying scrutiny where risk is identifiable rather than subjecting every user to additional checks, so that most participants never notice the controls at all.
He said the clearest warning signs include groups of accounts or wallets that appear connected, positions taken with unusual confidence just before an event, profits concentrated among a small number of participants, and odd win rates on long-shot outcomes. He also suggested monitoring should be calibrated to the nature of each contract, with military-related markets warranting closer attention than routine economic releases.
Fama said Sift's work focuses on spotting relationships between signals that look unremarkable in isolation, such as shared devices, linked identities, and trading patterns that appear too coordinated to be reviewed account by account but become visible in aggregate.
He acknowledged that artificial intelligence cuts both ways: the same tools that help detect coordinated fraud can also be used to generate synthetic identities and automate fraudulent activity at scale. Fama described it as an arms race, and said the aim should be continuous detection that informs contract design, access controls and market approval, rather than enforcement limited to individual accounts.