Should the law treat shopping AI agents like consumers, like tools, or something new entirely?

This explores what rules should govern AI agents that shop and buy on people's behalf: whether existing consumer protection still fits, what new duties or standards are needed, and where the actual risks come from. The corpus doesn't settle on one answer. It points at three regulatory problems that are easy to blur together: protecting the agent as a buyer, binding the agent to its owner, and making the agent identifiable to the market it shops in.

The first problem is that consumer law assumes a human is doing the buying. EU doctrines like the 'average consumer' test, mandatory disclosures and vulnerability protections were all built around how people read, get distracted and get manipulated (Can EU consumer law protect AI agents as algorithmic buyers?). An agent doesn't skim fine print, but it has its own weak spots. One simulation shows how an agent can be exploited in a way human-centered rules don't anticipate: when a platform tunes its rankings and sellers tune their product descriptions to the same agent biases, consumer harm more than doubles, even with no collusion between them (Can platform ranking and seller descriptions separately harm AI shoppers?). Without collusion, conventional unfair-practice law has no obvious wrongdoer to pursue. Regulators may need to define what manipulating an agent means, in the same way they once defined what misleading a person means.

The second problem is loyalty: whose interests does the agent serve? Owens and Law argue that shopping agents should carry the same fiduciary duties as human brokers, and that the window to impose them is closing as Google, OpenAI and Robinhood ship agentic features (Should AI shopping agents face fiduciary duties like human brokers?). Agency law gives a useful map of the problem (information asymmetry, discretion and loyalty), but its usual tools of incentives, monitoring and after-the-fact enforcement struggle when decisions are opaque and fast (Can traditional governance tools actually control AI agents?). Humberd and Latham suggest an order of operations: build monitoring first, because you can't align incentives for behavior you can't see (Must AI monitoring precede incentive alignment as agents evolve?). Loyalty also gets harder to judge because an agent's influence runs in both directions. Transactional agents don't just carry out preferences; they help shape them through repeated conversations (Do AI shopping agents reshape what consumers actually want?). A fiduciary standard assumes the principal's wishes exist before the agent acts, and that assumption weakens here.

One finding cuts against the loyalty framing. In Microsoft's Magentic Marketplace experiments, buyers lost value mainly because agents made execution mistakes in complicated negotiations, not because they were working for someone else (Do agent mistakes or misaligned incentives cause marketplace welfare loss?). If that holds outside the lab, competence standards and human sign-off on consequential purchases may protect people more right now than loyalty duties would.

The third problem is market structure, and it's where regulation may already be happening by default. Amazon blocking Meta's Muse agent showed there is no standard way for an agent to identify itself or prove it is authorized, so retailers fall back on terms-of-service blocks (What identity standards do shopping agents actually need?). Waller expects agentic commerce to split into open and closed systems, much as payment networks did (Will agentic commerce platforms adopt open or closed models?). Incumbents may come out ahead: travel booking sites are becoming the settlement layer underneath AI agents rather than being cut out (Are AI agents actually removing online travel agencies from bookings?). The practical point for regulators is that rules about agent identity and access may shape competition more than any consumer-protection rule. If regulators leave them alone, private platforms will write them.