The trajectory of agentic commerce is splitting. Europe is prioritizing a regulatory-first architecture, while the United States remains stalled by a lack of federal guidance and unresolved liability questions. The same underlying technology is producing two distinct commercial realities.
The European Regulatory Backbone #
Europe is utilizing the PSD3/PSR framework to define its deployment. The legislation, which assumes a human is at the keyboard, creates immediate friction. Institutions are pivoting from traditional Know Your Customer protocols to a new imperative: Know Your Agent. Despite this, the infrastructure is moving quickly.
Live end-to-end payments have been executed by Santander, Mastercard, ING, and Worldline. On July 2, 2026, ING, Worldline, and Visa completed an agentic payment in Germany utilizing Visa Payment Passkeys for biometric authentication, marking a distinct milestone in secure, automated transaction flows. Simultaneously, Mastercard has enabled all issuers in Europe at the network level for Agent Pay, supported by a new Lisbon Centre of Excellence for Innovation designed to accelerate these agent-driven payment experiences.
Kelly Devine, Mastercard Europe President, notes: “Agentic payments represent a profound shift in how commerce is initiated and executed. With Mastercard Agent Pay, we are applying the same principles that have defined our network for decades – security, trust, interoperability and global scale – to a new era of AI-enabled commerce.” Madalena Cascais Tomé of the Worldline Executive Committee adds: “Agentic commerce is no longer theoretical, it is production-ready today.”
The US Liability Vacuum #
The US market features a proliferation of agent platforms, yet merchant adoption is minimal. The issue is not technical capability, but liability. The US Treasury OIG has flagged ambiguity in Regulation E regarding agent authorization. Furthermore, the AI AGENT Act introduced in July 2026 focuses on fiduciary duties rather than the critical question of liability allocation for agent misexecution. In response to this regulatory silence, the Consumer Bankers Association has recommended that the industry develop its own private network rules to govern these interactions.
Stephanie Cohen, Cloudflare CSO, highlights the fundamental challenge: “The internet was built for human interactions, but the infrastructure of the future must be built for autonomous ones.”
Structural Irony #
Data from Hypertrade shows a 4,700% year-over-year increase in AI-generated traffic to retail sites, yet agentic commerce accounts for less than 1% of US e-commerce. If the infrastructure is ready, why is the merchant ecosystem so cautious? The numbers provide a stark answer. According to the Visa Earning Trust report, only 23% of US consumers trust generative AI to handle payment transactions. On the merchant side, PYMNTS Intelligence reports that 93% of merchants believe the AI provider should bear the financial loss for incorrect purchases.
The mismatch is clear: consumers are wary of the autonomy, and merchants are unwilling to absorb the risk of machine error. Until the liability question is resolved through either federal mandate or private network consensus, the US will likely continue to see this gap between technological potential and commercial reality.