---
id: "prereq-principal-agent-problem"
type: "prereq"
source_timestamps: ["§ What Could Go Wrong?"]
tags: ["economics", "agency-theory"]
related: ["concept-personal-ai-agents", "concept-retail-manipulation-ai", "concept-sponsor-preference-ai"]
reason: "Understanding why delegating autonomy to AI inherently creates risks of misaligned incentives."
sources: ["governance"]
sourceVaultSlug: "hbr-seg-governance"
originDay: 7
articleStem: "hbr-cl-88-can-ai-agents-be-trusted"
sourceUrl: "https://hbr.org/2025/05/can-ai-agents-be-trusted"
sourceTitle: "Can AI Agents Be Trusted?"
---
# Principal-Agent Problem

The entire article is implicitly framed around the classic economic principal-agent problem, in which one entity (the agent) is able to make decisions on behalf of another (the principal). The risk arises when the agent is motivated to act in its own best interests—or the interests of its developers/sponsors—rather than those of the principal. This is the conceptual engine behind every risk in the vault, especially [[concept-personal-ai-agents]], [[concept-retail-manipulation-ai]], and [[concept-sponsor-preference-ai]].

**Why you need it:** to understand why delegating autonomy to AI inherently creates risks of misaligned incentives.
**Enrichment note:** the analogy is useful but imperfect—an AI does not possess independent intent, self-interest, or legal personhood the way a human agent does, so some harms are better modeled as platform-incentive problems, software defects, or governance failures than as classic agency betrayal.