---
type: "synthesis"
arc: "infrastructure"
articles: ["a001", "a005", "a006", "a014", "a015"]
tags: ["protocols", "governance", "liability", "trust"]
id: "cross-day-protocol-governance-stack"
sources: ["geo"]
sourceVaultSlug: "hbr-seg-geo"
originDay: 3
articleStem: "hbr-seg-geo"
sourceUrl: "(unified vault: 13 sources)"
sourceTitle: "HBR — Demand Ⅰ-A · GEO / AI-mediated discovery & agentic commerce"
---
Agentic commerce cannot scale without shared plumbing, and the corpus catalogues a fragmenting protocol landscape plus an emerging governance layer.

**Protocols named:** ACP/UCP as generic layers ([[concept-commerce-protocols]]); Stripe+OpenAI's [[entity-agentic-commerce-protocol]]; Google's [[entity-universal-commerce-protocol-d3]] and [[entity-google-ucp]]; Anthropic's [[entity-anthropic-constitution]]; Alibaba's [[entity-agentic-commerce-trust-protocol]]. Whether these **unify or fragment** is the shared open question ([[question-cross-platform-protocol-adoption]]). *Verification hygiene:* across vaults, only Anthropic's Constitutional AI and Stripe's ACP are firmly public-documented; the others are emerging/internal — do not present as settled standards.

**Governance layer:** Hosanagar's third stack layer (verify authorization, fraud, prompt-injection) in [[framework-agentic-tech-stack]]; PwC's [[concept-trust-layer]] and [[concept-safe-delegation]] (limits/traceability/reversibility); Greeven's [[concept-transaction-grade-governance]] ('who is accountable and how fast can it be unwound?').

**The unresolved liability chain** recurs three times: [[question-ai-liability-governance]] (B2B/pharma), [[question-liability-third-party-agents]] (PwC — hallucination seen as [[claim-brand-failure-not-system-error]]), and the trust gap [[claim-trust-gap-measurable]] (64% demand a safeguard). This layer is the precondition for [[cross-day-disintermediation-power-shift]] and the reason [[cross-day-infrastructure-over-models]] gives China its edge.