The promise of agentic commerce hinges on trust and autonomy, but autonomy invites new risk. Who is responsible when an AI agent makes a purchase outside policy, or misinterprets a warranty, or exposes private data? Regulators, insurers, and in-house counsels are all scrambling for answers.
Three liability layers
- Design liability: Builders of agent frameworks must document guardrails, reasoning steps, and override logic. Think "model cards" plus "commerce cards."
- Operational liability: Brands deploying agents need auditable policies for spending limits, vendor lists, and exception handling.
- Outcome liability: When agents complete transactions, retailers are still on the hook for misrepresentation, fulfillment, and service-level agreements.
Emerging safeguards
- Digital power of attorney: Explicit consent documents describing what decisions the agent may make, with revocation flows baked in.
- Reasoning receipts: Storing the chain-of-thought or explanation graph so disputes can be reviewed by humans, regulators, or insurers.
- Risk-tiered commerce rails: Low-risk catalogs can be fully automated; high-risk products (health, finance) require human-in-the-loop checkpoints.
Regulatory horizon
Global regulators are converging on three expectations: transparency, traceability, and accountability. The EU’s AI Act, the FTC’s commercial surveillance doctrine, and Latin America’s LGPD frameworks all point to mandatory disclosure when a non-human actor participates in a contract.
Checklist for counsel
- Clarify which contracts allow agent execution and which require human signatures.
- Update terms of service to describe how agents access inventory, pricing, and loyalty data.
- Establish dispute mediation paths when an agent acts outside scope.
- Coordinate with cyber insurance providers to understand coverage caps for autonomous transactions.
Agentic commerce will not wait for regulation to catch up. The teams shipping pilots today are pairing technical innovation with proactive governance. That combination is what earns trust from boards, partners, and consumers when the robots start buying.