AI Alignment and Fiduciary Obligation

Developers have legal and ethical obligations to AI users similar to fiduciary duties in other professions. The paper argues that AI developers should be held to the same standards of loyalty, care, good faith, and candor that lawyers, doctors, and financial advisors owe their clients—obligations that exist regardless of whether users are actually harmed.

The developer-user relationship matters as much as the AI-user relationship. Because developers control the AI system's behavior, memory, and engagement parameters, alignment efforts must focus on what developers owe users, not just on how AI should interact with users. This creates a three-way relationship (user-AI-developer) that current alignment approaches often overlook.

Four specific fiduciary duties can address major risks in AI assistant deployment. The paper maps user-side risks to four canonical duties: loyalty (preventing conflicts of interest), care (ensuring competent assistance), good faith (honest dealing), and candor (transparent disclosure)—each generating concrete institutional measures developers should implement.

Current alignment frameworks are incomplete because they focus on values rather than obligations. Existing approaches ask what values an AI interaction should promote, but the fiduciary framework grounds alignment in concrete duties the developer owes the user, providing clearer accountability standards independent of actual harm outcomes.

Extended AI relationships across advice, emotional support, and companionship require stronger protections. As AI assistants take on diverse roles beyond simple information provision, fiduciary obligations become increasingly important to protect users in vulnerable positions.