Episode 41: When AI Agents Go Shopping

Episode summary

Consumers are starting to hand shopping tasks to AI agents that compare prices, choose products and complete checkout. In Episode 41, hosts Daniel Achterberg and Grace Whitcombe talk with Kenji Watanabe-Price and Sophie Laurent-Brecht about what the law says when software acts for a person, and what retailers, platforms and agent developers should do about it today.

The conversation covers agency and authority, who bears the loss when an agent buys the wrong thing, how consumer protection rules on disclosures and cancellation apply when no human sees the screen, and how the EU AI Act treats agents that interact with people. It runs about 32 minutes.

Chapter notes

• 00:00 Introduction. The hosts set up the problem with a simple example: an agent told to find the best price on a winter coat.
• 04:10 Who authorized the purchase? Kenji explains how contract law treats an agent's acts and why terms of service are doing a lot of work.
• 11:45 Allocating liability. Vendor contracts, indemnities and caps for agent errors, and what retailers should ask for.
• 18:30 Consumer protection. Grace discusses disclosures, returns, subscription traps and what regulators have signaled about dark patterns aimed at agents.
• 25:15 The EU view. Sophie covers transparency duties and how documentation expectations reach deployers.
• 29:40 What to do now. Three practical steps for legal teams before the holiday peak.

Hosts and guests

Hosts: Daniel Achterberg, Partner and Global Chair, AI Practice, San Francisco; Grace Whitcombe, Partner and Head of Consumer Protection, Chicago.

Guests: Kenji Watanabe-Price, Counsel, Technology Transactions, San Francisco; Sophie Laurent-Brecht, Counsel, Brussels.