Law firm Mayer Brown published an advisory on Tuesday, November 17, on who bears the cost when AI agents break rules, PYMNTS reported. Regulators are expected to look first to the company that deployed the agent, according to the advisory, and whether that company can recover losses from its supplier may depend on what the contract assigns to each party and what it can prove about how the agent acted.
The advisory gives a procurement example, in which a purchasing agent agrees with other buyers' agents to cap what they pay suppliers, raising antitrust exposure. In a retail example, a sales agent encourages New York shoppers to buy scooters that exceed the applicable speed threshold, and the retailer could face enforcement even if it argued the vendor should have built a safeguard.
The advisory suggests contracts specify who maintains jurisdiction-specific product restrictions, who tests that the agent blocks prohibited sales, who can suspend it and who covers losses if a safeguard fails. It names audit rights, verification tools and decision logs as key safeguards, covering instructions, configuration changes, approvals and transactions. It also flags disputes over repeated orders, vendor updates that reduce accuracy, privacy incidents and use of customer data for training, and recommends a simulated incident walk-through with the supplier before launch.
