AI in action · Investor protection
ESMA Says Prediction Markets Are Rife With Insider-Trading Risk, and Frontier AI Raises the Stakes for Market Infrastructure
The EU securities regulator's second 2026 risk report keeps securities markets and crypto at very high risk, singles out prediction markets for market abuse after wallets created hours before February's strike on Iran made about $1.2 million, and says market infrastructure faces increased risk from developments in frontier AI. Law firms reading the report in October argue that AI-generated signals and bots make the abuse harder to police.
The European Securities and Markets Authority has put prediction markets at the top of its list of places where market abuse is hardest to detect, and the reason is partly AI. Its second Trends, Risks and Vulnerabilities report of 2026, covering the six months to June 30, assesses securities markets and cryptoassets as being at very high risk and says that market manipulation and insider trading risks reach new levels in the context of prediction markets, the platforms on which users bet on events from elections to earnings. Legal analyses of the report published in October by Skadden, Ashurst and others have brought it back into the discussion as US regulators debate how to treat the same platforms.
ESMA's examples are specific. A batch of newly created wallets placed bets hours before the February 2026 strike on Iran and made about $1.2 million; a later case identified nine linked accounts with a 98% win rate on Iran-related bets. The regulator says the use of crypto rails makes insider trading, wash trading and coordinated manipulation harder to detect, because traders on blockchain-based platforms can operate pseudonymously and control multiple accounts, and that the largest risks sit on exactly those platforms.
The AI content of the report is about infrastructure rather than trading. ESMA writes that market infrastructures and services face increased risk because of developments in frontier artificial intelligence, a line that echoes the European Systemic Risk Board's July warning on cyber resilience and Christine Lagarde's October account of models completing every step of a simulated attack. The law-firm commentary goes further than the regulator: Ashurst argues that AI-generated trading signals and automated bots may worsen information asymmetries and facilitate coordinated trading on prediction platforms, and that AI-generated misinformation and coordinated online campaigns could distort the prices themselves.
What ESMA does not propose is new law. The report takes the view that existing rules on market abuse are sufficient and does not call for legislation on prediction markets or tokenised equities. The commentators' objection is jurisdictional: much of the activity happens on platforms outside the EU's regulatory perimeter, and a rule that is sufficient on paper is not much use against a pseudonymous wallet on a chain the regulator cannot subpoena. That gap, rather than any gap in the rules, is the finding worth remembering.
For traders the warning is practical. A prediction market in which nine linked accounts can win 98% of the time on one topic is a market in which the other side of those bets was systematically losing to someone with better information. The same thin, pseudonymous liquidity that makes these platforms interesting makes them the easiest place in finance to be the uninformed party, and the regulator that watches them most closely has just said so.
Sources: NEXT.io — ESMA warns prediction markets "rife" with insider trading risk (2026); Integrity Research — ESMA Warns that Prediction Markets are "Rife with Insider Trading" (2026); Ashurst — Prediction markets: ESMA flags growing risks, but can they be policed? (Oct 2026)