Mastercard and Danske Bank have completed Denmark's first fully autonomous AI agent payment on live card rails, marking a major shift toward agentic commerce. The transaction utilized Mastercard's Agentic Token and Payment Passkeys to secure and authorize the payment without exposing raw card credentials. While financial institutions scale these technologies, legal experts and global regulators warn that existing frameworks, such as Regulation E and PSD2, are unprepared to handle the liability and dispute challenges of machine-initiated transactions.
Mastercard's technical framework for agentic payments
- ▪Mastercard's programmable consent architecture allows consumers to set spending limits, permitted merchant categories, and expiry windows enforced at Mastercard's network authorization step
- ▪Mastercard's Agentic Token is a surrogate credential built on the EMV tokenization standard that links a token to an AI agent's registered identity and a consumer consent policy.
- ▪Mastercard Payment Passkeys use the FIDO2/WebAuthn public-key standard to authenticate transactions initiated by AI agents without requiring real-time human approval.
Agent Pay for Machines
- ▪More than 30 industry partners, including Stripe, Coinbase, Cloudflare, Adyen, and Checkout.com, joined Mastercard's Agent Pay for Machines at the launch of Agent Pay for Machines
- ▪Mastercard launched Agent Pay for Machines on June 10, 2026, to support high-velocity, machine-to-machine transactions with permissions initially recorded on Polygon, Solana, and Base.
Danske Bank autonomous payment pilot
- ▪The Danish autonomous payment transaction processed through Danske Bank's live card-issuing and authorization systems, which serve approximately 3.7 million personal banking customers in Denmark.
- ▪Mastercard and Danske Bank completed Denmark's first autonomous AI agent payment on September 21, 2026, using a live Mastercard on existing bank card rails.
Danske Bank AI strategy
- ▪Danske Bank extended its partnership with Amazon Web Services in June 2026 to adopt Amazon Bedrock and Amazon Bedrock AgentCore to scale generative AI services.
- ▪Danske Bank's Forward '28 strategy aims to establish the financial institution as a leading Nordic tech- and AI-driven bank.
Scaling agentic startups
- ▪Mastercard announced a 22-company agentic commerce and services cohort through its Start Path startup engagement program to evaluate and scale agentic technologies.
- ▪Mastercard evaluates startup founding teams, funding, technology, and global expansion potential to determine if they can scale through the company's ecosystem.
Agentic commerce adoption
- ▪Sabrina Tharani of Mastercard stated that the first mainstream adoption of agentic transactions will be low-value, high-frequency purchases where consumers prioritize speed and convenience.
- ▪Sabrina Tharani of Mastercard stated that consumers are comfortable delegating browsing and price comparisons to AI agents, but purchasing authority remains a higher bar.
Agentic commerce optimization
- ▪Max Sinclair co-founded Azoma.ai to help brands with agentic commerce optimization and generative engine optimization.
- ▪Agentic commerce optimization is the process of improving how products are discovered, understood, evaluated, and recommended by AI shopping agents.
Azoma shopping analysis
- ▪An observational analysis by Azoma.ai during Q2 2026 found that 73% of citations in Alexa for Shopping responses came from affiliate sites, 16% from earned media, and 11% from brand-owned content.
- ▪Azoma.ai's Q2 2026 observational analysis evaluated millions of prompts and responses across all brands and product categories on its platform.
Regulatory challenges for agentic payments
- ▪The Payment Services Directive 2 governs EU payment initiation services, but whether an AI agent acting for a consumer requires separate authorization under the directive remains an open regulatory question.
- ▪The Electronic Fund Transfer Act and Regulation E in the United States govern consumer protections for electronic fund transfers but were built around human-decisioned transactions rather than agentic payments.
- ▪The Bank of England, the Financial Conduct Authority, and the International Monetary Fund flagged regulatory gaps in agentic payments in 2026.
Debatable claims
- ▪Banks should bear the liability when AI agents exceed their delegated purchasing authority
- ▪Programmable consent limits are insufficient to protect consumers from AI payment fraud
- ▪Traditional bank card rails are the wrong infrastructure for high-velocity machine transactions
- ▪Existing financial regulations are inadequate to govern autonomous AI agent payments
Story comments
Loading comments…