Agent-to-Agent Risk Across Company Boundaries: The Next Governance Gap (Oct 2026)

Agent-to-Agent Risk Across Company Boundaries: The Next Governance Gap (Oct 2026)

The newest adoption barrier isn't inside the enterprise — it's the boundary between two companies' agents. A detailed analysis by Allan Dabre (senior manager of technology compliance at PwC) in Risk Management Magazine (Oct 6, 2026) frames the failure mode with a concrete scenario: "A manufacturer's demand-planning AI agent detects an impending shortage and reaches out, autonomously, to a supplier's fulfillment AI agent to expedite a shipment... The supplier's agent complies with its own rules, rerouting inventory allocated to a different customer... Neither agent malfunctioned. Each stayed inside the rules its own company gave it." The crux: "It is a question of operational control at the boundary between two companies' automated systems." (Risk Management Magazine)

The quantified backdrop, as assembled in the article:

• Deloitte's 2026 State of AI in the Enterprise survey found "only 21% of companies have a mature governance model for autonomous agents" — so most organizations wiring agents across supply-chain relationships are doing so without governance built for cross-boundary complexity.
• Gartner projects 40% of enterprise applications will ship with task-specific AI agents by end of 2026, up from under 5% in 2025, and that AI agents will intermediate more than $15 trillion in B2B spending by 2028.
• OWASP's Top 10 for Agentic Applications 2026 names supply-chain risk, inter-agent communication, and cascading failures among its highest-priority categories — "a standards body categorizing this failure mode before the first major public incident is itself worth noting."

Why existing controls don't stretch. Every standard agentic-governance control (know what the agent does, restrict it, log it) "assumes the agent being governed operates inside a perimeter the company itself controls¹." The article identifies three cross-boundary failure modes — identity/authorization ambiguity (is the counterparty signal really from an authorized agent?), cascading action without a shared circuit breaker (each link looks fine locally; the combined outcome is the problem), and a fractured audit trail (neither side holds the full exchange record). Legally, EU AI Act Article 26's oversight/log requirements "was written for a single company deploying a system it controls end to end," and a June 2026 Berkeley Technology Law Journal analysis found that once an agent invokes a tool or another agent at runtime, "liability disperses among model providers, system providers, deployers and tool providers, with no single actor holding full visibility." Insurance compounds it: new ISO endorsements effective 2026 let carriers exclude generative-AI claims from standard GL policies, with similar exclusions filed across E&O and D&O lines — an agent-to-agent loss may fall into exactly that gap.

Proposed boundary controls: mutual agent authentication for cross-organizational exchanges; escalation logic keyed to cross-boundary action categories (human review whenever an action executes as the downstream result of a counterparty's agent signal); a shared, interoperable agent-to-agent logging standard agreed before go-live; and treating a partner's agentic-governance maturity as an ongoing due-diligence question rather than a point-in-time vendor questionnaire.

What it means for adoption: the efficiency case for agents is strongest precisely at the cross-company seam (procurement, logistics, demand signaling), and that is where governance, law, and insurance are thinnest. Expect contract terms and shared logging standards to become procurement blockers in 2027. This extends the accountability-vacuum picture in Governance and Security: Senate "Rogue AI" Hearing, METR Testimony, and the Kill-Switch Gap (Early October 2026) and sharpens the "governance lags deployment" half of The Enterprise AI Agent Production Gap: Gartner's 17% Deployment / 40% Cancellation Math (October 2026 Wave).