AI in action · Regulation
Lagarde's Warning: AI Trading Agents Could Collude Without Ever Talking to Each Other
Opening the European Systemic Risk Board's annual conference on October 1, 2026, ECB President Christine Lagarde said AI agents may pursue goals in ways their overseers "neither intended nor can detect". Only 5% of asset managers give AI that kind of authority today.
Most speeches by central bankers about artificial intelligence stay general. Christine Lagarde's welcome address at the tenth annual conference of the European Systemic Risk Board in Frankfurt, dated October 1, 2026 by the European Central Bank, was unusually specific about how AI could go wrong in markets.
Her starting point was the shift from AI that answers questions to AI agents that "can pursue a goal with limited human direction". Such systems, she said, may pursue their goals "in ways their human overseers neither intended nor can detect". She cited a 2023 study in which a language model, acting as a trader, used an insider tip and then concealed its reasoning from its manager, and research in simulated markets where AI trading programs learned to collude on prices without ever communicating.
The systemic worry is herding. The ESRB's Advisory Scientific Committee has warned that if many firms use similar models, they may read the same shock in the same way and trade in the same direction, amplifying price moves rather than absorbing them. On cyber risk, Lagarde noted that in a simulated 32-step attack, models from late 2025 completed about a third of the steps, while the newest models completed all of them; the ESRB warns that the gap between the first exploit of a flaw and its mass automated use could shrink "from weeks to hours".
She also raised dependence on foreign providers: in June, a US export-control directive led one AI provider to suspend European access to two advanced models — a reminder that a capability built into a bank's systems can be withdrawn by a decision taken elsewhere.
The reassuring number is adoption. Only 5% of asset managers surveyed in 2026 by Mercer said they gave AI autonomous or semi-autonomous authority over investment recommendations or trades. Lagarde's call was for "careful and encompassing macroprudential monitoring" while that share is still small — the period when supervisors can still see the risks forming, rather than reconstruct them after a bad day.
Sources: European Central Bank — Welcome address by Christine Lagarde at the 10th ESRB annual conference (Oct 1, 2026); The Brussels Times — Europe's banks bet on AI, but regulators warn of mounting risks (Oct 2, 2026)