A two-tier trust architecture for AI-driven commerce is taking shape in a single day's announcements: at the transaction layer, Visa, Mastercard, and Ant International are converging on a shared Know-Your-Agent identity framework, while at the infrastructure layer, Akamai and MuleSoft are wiring runtime API security directly into the orchestration plane that governs agents. Together, these moves sketch the plumbing that payment networks and enterprise platforms believe must exist before autonomous software can safely spend money at scale.
Why it matters
McKinsey's QuantumBlack research unit projects that AI agents could mediate $3 trillion to $5 trillion of global consumer commerce by 20306. That figure, cited in the Ant-Visa-Mastercard announcement itself, is doing real work: it is the shared justification for three competitors to begin aligning verification standards before any single network captures the market1,5. The implication is that none of the three believes it can impose a proprietary trust layer on agentic commerce alone, and all three would rather co-define the rules than cede that role to a platform company or regulator.
The big picture
The Know-Your-Agent framework announced September 10 rests on three pillars: operator traceability, which links each agent to a validated operator, cardholder, or business; shared certification, which assesses each agent against security and behavioral requirements; and continuous monitoring using identity and transaction signals4. Each company already operates its own protocol: Visa's Trusted Agent Protocol, Mastercard's Verifiable Intent, and Ant International's Agentic Mobile Protocol13. The new effort will explore how those separate approaches can work toward common principles rather than replacing any network's individual verification process11.
The collaboration will run through BuildFin.ai, an industry platform convened by the Monetary Authority of Singapore that brings together financial institutions, technology providers, and researchers to develop responsible AI solutions for financial services10. Singapore's regulatory convening role positions the framework as a multilateral effort from inception rather than a bilateral deal between Western card networks.
One layer down, Akamai and MuleSoft announced an expanded integration connecting Akamai API Security with MuleSoft's Agent Fabric2,7. The integration establishes a bidirectional data synchronization layer: MuleSoft Exchange automatically synchronizes API specifications and environment data with Akamai, while Akamai feeds behavioral analytics and threat-risk scores back into MuleSoft's unified control plane3. More than 20 joint customers are already using the integration. The urgency is quantified: 87% of organizations reported an API-related security incident during 2025, according to the 2026 Akamai API Security Impact Study.
ANALYSIS The two announcements address different failure modes of the same problem. The KYA framework answers the question "who is this agent and is it authorized to transact?" The Akamai-MuleSoft integration answers "what is this agent actually doing at runtime, and does it match policy?" Neither layer is sufficient alone. An agent could pass identity verification yet exploit an undocumented API endpoint; conversely, runtime monitoring is less useful without a verified identity to attribute anomalous behavior to.
Ant International's chief innovation officer, Jiang-Ming Yang, said the companies look forward to "expanding collaboration as the industry draws on richer signals to enhance trust and simplify transactions"9. The phrase "richer signals" points toward a future in which identity verification and runtime telemetry merge into a continuous trust score, not a one-time gate.
The Akamai-MuleSoft side makes the infrastructure dependency explicit. As enterprises deploy AI agents at scale, Agent Fabric governs the underlying agents, APIs, and MCP servers that make agentic workflows possible. Planned updates include native MuleSoft onboarding features and expanded runtime security specifically for AI and MCP environments. MCP servers are becoming a critical control surface: they sit between agents and enterprise data, and without visibility they become vectors for shadow AI and unmonitored data exfiltration, as Akamai's own framing acknowledges.
The companies said the KYA initiative could reduce integration complexity and duplicative agent identity verification while maintaining robust trust and risk controls. It could also enable faster time-to-market and lower integration cost for new agentic services. The economic logic is straightforward: if an agent registered with Visa must repeat the entire verification process with Mastercard and again with Ant, the friction tax on agentic commerce grows with every network it touches. Interoperability at the identity layer removes that multiplicative cost.
What's next
The KYA work will proceed through BuildFin.ai, which builds on the Monetary Authority of Singapore's Safeguards for Agentic Finance at Runtime framework and initially covers payment ecosystems in Singapore. On the infrastructure side, Akamai and MuleSoft have flagged expanded runtime security for AI and MCP environments as a planned update. The open question is whether the KYA framework remains a set of common principles that each network implements independently, or whether it hardens into a shared technical standard with enforceable compliance requirements. The answer will likely depend on whether the 87% API-incident rate drops fast enough to keep regulators from mandating something more prescriptive.