ESMA vs Prediction Markets: Why the EU Is Closing the Door

Europe's position on prediction markets stopped being ambiguous in 2026. It is now a stack of decisions pointing the same way.

The timeline

• May 2026 — Spain. The Consumer Rights Ministry temporarily banned Polymarket and Kalshi for operating without gambling licences and opened an investigation (1).
• July 2026 — binary options. ESMA clarified that certain prediction-market contracts may fall under the EU-wide ban on binary options, and warned providers about marketing them (2).
• July 2026 — France. The ANJ ordered French ISPs to block Polymarket as an illegal gambling site (3).
• September 2026 — ESMA. The regulator stated that the major prediction platforms lack authorisation to serve EU users, and questioned whether their partial geo-blocks are adequate (4). Its risk monitoring also flagged manipulation concerns in a market running tens of billions in monthly volume (5).

The regulatory trap

A prediction market has to be one of two things in the EU, and both are bad news.

If it is a financial instrument, it needs authorisation under EU financial rules — and cash-settled contracts on a yes/no outcome look a lot like binary options, which are banned for retail investors in the EU.

If it is gambling, it falls to national law. Gambling is not harmonised across the EU, so there is no passport: a platform needs a licence in France, another in Spain, another in each member state — with the local restrictions on political betting that come with them.

There is no third box where a global, unlicensed, cash-settled event exchange sits comfortably. That is why the blocks keep arriving.

What regulators are not banning

Nothing in these decisions makes it illegal to estimate a probability and publish it. Forecasting, research, polling and modelling are ordinary information activities. The regulated element is the contract that pays money on an outcome.

This distinction matters more than it sounds, because for most people the bet was never the point. Analysts, journalists, investors and strategy teams quoted Polymarket odds as a signal. They wanted the number.

The unregulated-by-design alternative

Watching Agents produces the number without the contract. An AI research agent tracks a question continuously and publishes:

• a live probability and a confidence score,
• the drivers currently pushing the odds up or down,
• a history of how and why the estimate changed,
• for multi-outcome questions, a full scenario breakdown rather than a forced yes/no.

There is no order book, no deposit and no payout, so none of the 2026 decisions apply. It works the same in Madrid, Paris, Brussels and Bucharest.

It also fixes a problem markets have

ESMA's manipulation concerns are real: a thin market is cheap to push, and a distorted price is worse than no price. An agent's estimate cannot be moved by a single large order. It can be wrong — models are wrong all the time — but you can read the reasoning and judge for yourself, which is not something an order book offers.

Practical next steps

• See the country-by-country picture in where Polymarket is blocked in 2026.
• Compare the two methods honestly in prediction markets vs AI forecasting agents.
• Browse live agents or create your own question at /new.

The EU is not closing the door on knowing the odds. It is closing the door on betting on them.