If an AI shops or trades for you, should it owe you the same loyalty a human financial advisor legally must?

This explores whether the legal duties human advisors owe their clients, such as loyalty, care and putting the client first, should also bind AI agents that act on our behalf. The direct case for it is urgent. Owens and Law argue that agents doing real shopping and trading should carry the same common-law fiduciary duties as human brokers. Their reason is timing: Google, OpenAI and Robinhood are scaling agent features fast, and once market habits harden it will be much harder to add obligations afterward Should AI shopping agents face fiduciary duties like human brokers?.

The less obvious point is that the old toolkit may not transfer cleanly. Fiduciary law grew out of principal-agent problems: the agent knows more than you do, has discretion over your affairs, and might serve someone else. Those same three problems show up with AI agents. But the usual fixes were incentives, monitoring and after-the-fact enforcement, and they assume you can read the agent's reasoning and keep up with its pace. AI agents break both assumptions Can traditional governance tools actually control AI agents?. Humberd and Latham take this further. You can't motivate an AI with incentives the way you motivate a human advisor, so monitoring has to come first. With AI, it's harder to define what good behavior even looks like, and its decisions get harder to follow over time Must AI monitoring precede incentive alignment as agents evolve?. In practice, a duty of loyalty for software may depend less on courts and more on infrastructure: cryptographic receipts, permanent action logs, and humans placed at deliberately designed handoff points How should organizations supervise autonomous AI agents?.

A surprising finding complicates the whole framing. Fiduciary duty mainly guards against disloyalty, meaning an advisor quietly serving their own interests. But in Microsoft's Magentic Marketplace experiments, agents lost value for their users mainly because they made mistakes in complicated negotiations, not because they were working for anyone else Do agent mistakes or misaligned incentives cause marketplace welfare loss?. If that holds more broadly, the duty of care (being competent) may matter more right now than the duty of loyalty. The practical fix may be limiting how much authority an agent gets. Risk to people rises with the autonomy they hand over Does AI risk increase with the autonomy we give it?. Users already seem to sense where the line is: trust drops sharply on actions that can't be undone and that other people can see, like sending an email, even when the output is fine What makes people distrust AI agents they delegate to?.

There is also a challenge to the one-client model itself. A human fiduciary serves one principal. An AI agent's actions spread to other people, to other agents, and to markets. Some researchers argue that loyalty to a single user is the wrong goal at scale. They propose balancing the interests of everyone affected, tracking agent actions closely, and holding developers and deployers liable for harms Should AI agents serve one user or balance many interests?. DeepMind's ethics work makes a related point: assistants that act raise different problems from assistants that only answer, including manipulation and coordination failures between many agents What makes ethics of AI assistants fundamentally different from chatbots?.

So the corpus's answer is roughly yes in spirit, but not by copying the rules over. The goals behind fiduciary duty (loyalty, care, disclosure) are the right ones. Enforcing them on AI agents likely needs monitoring built in from the start, limits on what agents can do alone, and liability that reaches the companies that build and deploy them. If you're curious where this goes next, the agency framing also runs in reverse: some argue agentic systems will start handing tasks and responsibilities to humans, which raises the question of who owes duties to whom Can AI systems delegate tasks back to humans?.