After Comet, the liability question for agents shifts from the vendor to the user
The Ninth Circuit's finding that users rather than Perplexity 'accessed' Amazon establishes the principal-instrument framing in American law. It resolves the immediate threat to consumer agents and opens a harder question: what a user is responsible for when they cannot fully predict what their agent will do.
Treating the agent as an instrument is doctrinally tidy and matches how the law handles other tools. It also imports the assumption that the person wielding the tool understands and directs it — an assumption that fits a hammer better than a system which plans its own multi-step route to a goal.
The gap will be tested quickly. A user who instructs an agent to find the best price has not authorised any particular sequence of requests, and the agent may take actions the user would not have chosen and could not have anticipated. Principal liability without meaningful foresight is a doctrine under strain.
For now the ruling supplies the certainty the market needed: agents can transact on public sites without their vendors facing hacking claims. Sites keep contractual and technical remedies. But the deeper allocation — how responsibility divides between a person, a vendor and a system that improvises — remains open, and this is the first case that forces the question.
Radical Data Science — AI news briefs bulletin board for August 2026 → · AI Weekly — AI news today, August 5 →