U.S. Congress Holds First Hearing on AI Agent Payment Rules: Authorization, Settlement, and Identity Take Center Stage

The U.S. House Financial Services Committee held a hearing on Wednesday (September 2, 2026), marking the first systematic effort to bring AI-agent-driven payments into the regulatory fold. Circle President Heath Tarbert, alongside executives from the Florida Bankers Association and the New York Stock Exchange, sparred over authorization mechanisms, smart contract settlement, and identity verification when software agents initiate transactions. The hearing signals that U.S. lawmakers are beginning to confront a question that has largely lingered in industry white papers: when machines spend money on behalf of humans, does the existing payment legal framework still apply?

In his testimony, Tarbert framed the shift as a migration at the infrastructure level. "We are seeing the plumbing of the financial system change because of digital assets," he said, describing Circle's vision as building an "internet financial system." He further emphasized AI's accelerant role: "Software agents transact on behalf of individuals and businesses. They don't stop for nights, weekends, or holidays. They need fast, programmable, always-on rails." His remarks directly targeted the physical limitations of traditional payment rails—business hours, manual reviews, and batch processing—where AI agents make decisions in milliseconds while bank clearing often takes days.

The core tension of the hearing centered on the fact that current law assumes "humans" are the transacting parties. KYC rules, fraud liability, and sanctions screening are all designed around natural person identities. When transaction instructions are generated by algorithms, how do banks confirm who is authorizing? Circle argued that smart contracts can embed authorization logic, but lawmakers pressed: if contract code contains vulnerabilities or is maliciously exploited, how is liability allocated? The NYSE's mention of tokenization initiatives and the banking sector's interest in tokenized deposits hinted that Wall Street is exploring multiple regulated pathways for moving funds on blockchain—all of which depend on a clear definition of the "originator."

For the payments industry, the hearing's significance is more symbolic than immediately legislative. It signals that regulators are beginning to treat AI agent payments as a distinct issue, rather than a simple extension of existing electronic payment rules. From OceanAlt's perspective on agent payment compliance and security, several concrete issues were put on the table. First, how can pre-settlement screening adapt to machine speed—traditional sanctions list scans may require second-level responses, while agent transactions could complete in milliseconds? Second, the absence of KYA (Know-Your-Agent) standards means banks cannot assess the "creditworthiness" of a software agent the way they vet human customers. Third, the digital expression of authorization intent—whether per-transaction limits, daily cumulative caps, and beneficiary whitelists can be encoded into smart contracts with legal effect.

The hearing did not produce specific legislation, but it revealed cross-party consensus: AI agent payments cannot operate outside anti-money laundering (AML) and sanctions compliance frameworks. Tarbert's testimony suggested Circle is willing to participate in regulatory sandbox testing, while NYSE's tokenization plans may first take root within regulated exchanges. For stablecoin issuers, payment service providers (PSPs), and banks, this means compliance cost structures will shift—future due diligence may need to satisfy requirements for both human and machine clients.

Also noteworthy is the timing: the hearing took place in September 2026, during the window when U.S. stablecoin legislation (such as the GENIUS Act) is advancing. Lawmakers discussed AI agent issues alongside tokenized securities, suggesting they are attempting to address both "who owns assets" and "who authorizes transfers" within a single framework. If subsequent legislation writes KYA requirements into stablecoin regulation, the adoption of machine payment protocols like x402 will be directly influenced by compliance standards—the protocols themselves do not solve identity, but regulation will force them to integrate identity layers.

For global markets, the U.S. stance carries barometric significance. The EU's MiCA framework has yet to define AI agent payment provisions, while the UK FCA's sandbox is just getting started. If the U.S. takes the lead in defining "KYC alternatives for software agents," it could serve as a reference for international standard-setting bodies like the Bank for International Settlements (BIS) when crafting cross-border norms. But the hearing also exposed divisions: Republican lawmakers focused more on innovation space, while Democrats emphasized consumer protection—a tension that will play out repeatedly in subsequent legislation.

For now, what the industry can be certain of is that AI agent payments have moved from proof-of-concept to regulatory scrutiny. Circle's "always-on rails" advocacy contrasts with the banking sector's caution, while NYSE's tokenization plans offer a third path—establishing clearinghouses for tokenized assets within existing exchange rules. Whichever path prevails, pre-settlement firewalls and identity verification will shift from optional configurations to mandatory requirements. The hearing did not provide answers, but it advanced the question from "can it be done" to "how to do it compliantly."

---

Original source: PYMNTS · https://www.pymnts.com/news/payments-innovation/2026/lawmakers-test-the-rules-for-ai-agents-that-move-money/