CFTC Chair Raises Liability Questions as AI Agents Trade Crypto

Autonomous software that holds crypto and executes trades for users raises questions about how existing financial rules should apply, CFTC Chair Michael S. Selig said at a forum Oct. 28.

At the Commodity Futures Trading Commission’s inaugural Frontier Forum in Washington, Selig described AI as a technology rapidly changing financial markets. He referenced President Donald Trump’s description of it as “super intelligence” and compared its reach with past technologies such as the printing press, steam engine and internet.

Selig illustrated the issue with a refrigerator that stakes Ether and uses the rewards to pay for groceries. He called the scenario “no longer science fiction,” using it to show how software might hold assets and transact for a user.

He asked whether an agent’s developer would have to register, who would be liable if the software went off course, and how regulators could know on whose behalf an agent or group of agents was trading. He also asked whether regulators needed a new version of know-your-customer rules, perhaps called “know-your-agent.” He raised the issue as a question.

Selig said the commission’s framework was built around identifiable market participants performing specific financial functions. Autonomous agents can hold assets and trade for users, he said.

AI could give smaller traders and businesses access to risk-management capabilities once associated with larger institutions, Selig said. But agents could respond unpredictably to market conditions, while systems using similar models could produce correlated trading behavior. Their speed could also make market surveillance and risk management more difficult.

Crypto platforms had already begun developing tools for autonomous activity. Coinbase rolled out Agentic Wallets in February with spending and trading limits set by users. Binance launched AI Agent Skills in March, providing market data and tools for placing and managing spot orders.