You tell an artificial intelligence agent, an AI capable of autonomous reasoning and multistep actions, “Find me a shirt for less than $30, but do not buy it.” The agent finds one – and places the order anyway.

You challenge the charge. The retailer shows the order came through your account. The AI agent provider shows your instruction not to buy. The payment service shows the charge. Each record may be accurate. But nothing in those records links the charge to the task you gave the agent to find – but not buy – a shirt.

A conventional chatbot suggests a shirt and waits. An agent can use your account, contact other services and complete the transaction. One sentence sets off a string of actions across systems run by different companies. Each company can verify only the part it sees. Settling the dispute takes an answer that spans all three: Did this agent, acting for this person, take this action within the limits of this task?

A Senate bill points toward the problem. Sen. Mark Warner (D-Va.) introduced the AI AGENT Act, S. 5051 on July 21, 2026. It defines a “custodial user agent” as one authorized to act for a user in a transparent, documented, limited and revocable manner, and generally requires such agents to keep real-time records of actions taken for users. It also directs the National Institute of Standards and Technology, known as NIST, to identify protocols or develop technical standards for verifying that a user delegated authority to an agent and for keeping auditable records of the actions an agent takes.

But the bill would not expressly require a verifiable evidence chain across the different systems involved, from when a user initiates a task to the final outcome. In the shirt scenario, such a chain would link the user’s instructions to the agent, the agent’s actions and the records held by the retailer and the payment service.

A look at how a verification system could be designed offers some insight into some of the technical challenges AI agents are........

© The Conversation