Visa, Mastercard and Ant International agree to honor each other's AI agent registrations

The concrete promise is narrow, and worth stating in the currency a builder cares about: register once. Jiang-Ming Yang, Ant International's chief information officer, told CNBC that an agent registering with Ant would not need to register again with Visa or Mastercard ^[9]. That is credential recognition across registries, not a merged protocol; each network keeps its own, and the three are Visa's Trusted Agent Protocol, Mastercard's Verifiable Intent and Ant's Agentic Mobile Protocol ^[3].

The three named workstreams say where the engineering burden lands. Cross-network traceability of operators and shared certification are onboarding gates; continuous transaction monitoring is a runtime obligation ^[2]. An agent that clears certification is still being observed afterward, which makes logging and intent attestation load-bearing choices in the first pilot rather than a later hardening pass.

None of the protocols being reconciled is old. TAP launched in October 2025 and has 12 partners, among them Adyen, Shopify and Stripe ^[5]. Mastercard introduced Verifiable Intent in March 2026, open-source and co-developed with Google ^[6]. Ant released AMP in April 2026 ^[7]. At the September announcement they were roughly eleven, six and five months old ^[18]. Aligning specifications that early is far cheaper than retrofitting them later, and it also means very little production volume has passed through any of them.

The one adoption figure on offer is a 4,700% surge in AI-driven traffic to US retail sites, which the Forkast account also calls largely experimental ^[14]. A percentage change with no stated base gives a direction and not a magnitude, and traffic is not checkout.

The demand side is measured more usefully. In the Product.ai Trust in AI Commerce Report from April 2026, 14% of consumers said they trust AI to execute purchases without verification ^[12], leaving 86% who want a check somewhere in the loop ^[17], and 42% said they will not trust AI for purchases above $25 ^[13]. Rubail Birwadker of Visa argues that without trusted identity and explicit permissioning, agents cannot participate in commerce at scale ^[11]. Those survey numbers are consistent with identity being necessary and well short of sufficient.

Set the ambition against a real base. McKinsey projects $3 trillion to $5 trillion of global consumer commerce orchestrated by AI agents by 2030 ^[8]. Worldpay put digital wallet spending above $13 trillion in 2025, 56% of global e-commerce value and 33% of point-of-sale value ^[7]. The McKinsey range is 23% to 38% of that wallet figure ^[19]; the two counts do not share a denominator, so it is not a share, but it fixes the order of magnitude being asked for. Ant's framing of a Permission Gap treats missing trust infrastructure as the constraint ^[16], while the consumer figures point at a scarcer input: a shopper willing to let an agent spend more than $25 unsupervised.

Mastercard's Pablo Fourez describes the goal as a consistent way for merchants, platforms, wallets and issuers to recognise trusted agents and verify that actions reflect the user's intent ^[10]. Whether that arrangement stays open or hardens into a gated ecosystem favoring the incumbents is, on the Forkast reading, the live risk ^[20].