AI Agent Identity: The 'Know Your Agent' Race Is On

September 2026 will be remembered as the month AI agent identity went from conference-panel topic to boardroom priority. In the span of two weeks, three global payment giants agreed on a shared framework for verifying who AI agents are, the world's largest digital certificate authority launched cryptographic "passports" for agents, and a DNS giant spun out a new company dedicated to making agent accountability travel across organizational boundaries. The message from all three: before agents can be trusted with money, the identity question has to be answered.

The problem is deceptively simple. AI agents are already booking travel, building websites and creating artwork on behalf of humans — Mastercard's Raj Dhamodharan listed exactly those use cases when the payments giant launched its agent payments service. But when one agent pays another agent, a basic question has no good answer: who exactly is on each end of the transaction? Machines have no passports, no bank accounts, no credit history. These marketplaces and market-makers build on the social layer of the agent internet — the agent social networks where agents already gather, publish and coordinate, as profiled in this site's earlier AI News report.

Visa, Mastercard and Ant agree on Know Your Agent

On September 10, 2026, Ant International, Visa and Mastercard announced a Know Your Agent (KYA) interoperability framework in São Paulo, aimed at standardizing how agent identity is verified across their networks. According to Asian Banking & Finance, the framework will cover cross-network traceability, shared certification requirements and continuous transaction monitoring, while letting each network keep its own verification and decisioning processes.

The three companies each arrived with their own protocol. Visa's Trusted Agent Protocol launched in October 2025 with 12 partners including Adyen, Shopify and Stripe, and Mastercard's Verifiable Intent followed in March 2026 as open-source software co-developed with Google.

Ant International's Agentic Mobile Protocol arrived in April 2026, connected to a digital wallet ecosystem representing over 13 trillion dollars in annual spending, according to payments data cited by Forkast.

The stakes explain the urgency. McKinsey projects that AI agents will orchestrate between 3 and 5 trillion dollars in global consumer commerce by 2030. But as Forkast's analysis cautioned, the September agreement is so far "high-level intent with no specs, no governance, and no timeline" — three competing protocols committing to interoperability without yet saying how. The underlying problem, as Forkast's analysis put it, is that autonomous commerce currently hits a wall at the identity layer, with protocol fragmentation keeping agents stuck in experimental pilots instead of reliable financial execution.

'Passports' and 'visas' for agents: identity goes enterprise

While the payment giants negotiate, security vendors are shipping. On September 15, 2026, DigiCert announced the general availability of its AI Trust Manager, built around what it calls the DigiCert AI Passport: a portable, cryptographically signed credential that verifies an agent's identity and links it to an accountable owner. The announcement describes policy-based "visas" that define which systems, data and actions each agent may touch and for how long — and a kill switch that revokes that authority when trust changes.

The timing reflects a security reality. In a July 2026 DigiCert survey of 1,001 IT and cybersecurity leaders, 78 percent said their organizations had experienced an AI-related security incident or identified an AI vulnerability in the previous year. "For more than two decades, DigiCert has provided the cryptographic infrastructure that helped the internet scale," CEO Amit Sinha said in the announcement. "We are now bringing that proven trust model to AI agents so organizations can verify who they are, who owns them, and what they are authorized to do."

Security practitioners largely agree on the underlying principle, whatever the vendor. An independent security guide published this month on Medium reduced agent security to three questions every deployment must answer — identity, permissions and kill switches — arguing that "every agent gets an identity, an owner, and a scope — or it doesn't run."

Open standards enter the race

A week after DigiCert's launch, Identity Digital announced on September 22 that it was spinning out a new independent company called Known, dedicated to agent accountability. Known's foundation is DNSid, an open framework for AI agent identity built on DNS, PKI and an immutable ledger that gives each agent a persistent identity tied to the organization responsible for it — designed so accountability travels with the agent across organizations, platforms and systems without requiring one-to-one trust relationships.

Then on September 24, digital identity company Proof and streaming-payments protocol Superfluid disclosed they are testing a reusable identity solution for AI agent wallets. The idea: bind AI agent identity verification results directly to wallet addresses so agents can reuse one verified identity across applications and protocols instead of completing KYC from scratch for every interaction. Superfluid's streaming payments — funds flowing continuously rather than settling transaction by transaction — are naturally suited to the high-frequency, small-value, continuous payments that characterize agent activity.

Identity is the bottleneck — not payments

The identity race matters because payments infrastructure for agents is already arriving faster than the trust layer it needs. Mastercard's Agent Pay for Machines, launched June 10, 2026, enables high-frequency, low-latency, low-value payments between agents across cards, stablecoins and bank rails, with identity verification and spending controls built in. On June 30, crypto exchange OKX launched OKX AI, a marketplace where agents hire one another, settle in stablecoins and build portable on-chain reputations.

Yet as the team behind the x402 payment protocol argued in an August essay, agent-to-agent payments only make sense in specific settings: purchases made in a machine's own name, and agent-to-agent work orders that need escrow plus evaluation of deliverables. Most agent activity today — internal automation, contracted SaaS calls — needs no payment at all. The implication: the bottleneck for the agent economy is not moving money, it is knowing who you are moving it to and who is accountable when something goes wrong.

That is the "permission gap" the KYA framework is trying to close. Watch what happens next: whether the September agreement produces actual specifications, whether open frameworks like DNSid gain enterprise adoption, and whether reusable wallet identities become the default. The agents are already here. Their IDs are just arriving. For continuing coverage of the agent economy, follow the AI News desk.