Invest1 publisher3 min readPublished
Christopher Waller flags AI payment agents for possible new rules
Fed Governor Christopher Waller, who oversees the Fed's payments operations, told Sibos 2026 that AI in payments, agents included, could require new rules. His remarks stop short of a proposal, so firms building agent payment flows have to judge where supervisors will look first.
The Investor · Invest desk
What happened
- Waller said agentic transactions could materially change the frequency and timing of payments, in a Tuesday speech in Miami at SWIFT's Sibos conference.
- He listed retail purchases by consumers or authorized agents on their behalf, business-to-business payments, and machine-to-machine micropayments for database queries.
- He said research shows large language models can significantly reduce false-positive alerts in sanctions and anti-money-laundering screening.
- He called concerns about more frequent and more sophisticated AI-enabled cyber attacks well founded and said the balance favors attackers over defenders.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- exposure Fintechs whose agents pay on a consumer's behalf hold the authorization record, so any rule on agent liability would be tested against logs those firms keep today.
- precedent A Fed governor endorsing research on LLM false-positive cuts gives banks a public reference point when they ask examiners to accept AI-based sanctions screening.
- constraint Waller's call for the industry to move deliberately on AI defense sets an expectation examiners can apply to payment operators before any agent rule exists.
The third form on Waller's list, micropayments that agents incur each time they query an online database, is the one I think strains current payment economics [6][1]. If one agent task triggers many paid queries, the count of payments rises faster than the money moved. Every fixed per-payment cost, such as sanctions screening or fraud scoring, is then spread over a smaller ticket.
Two phrasings in the same report differ in strength. Waller said the rise of AI in payments could necessitate new regulations [3]. American Banker also reported that he called LLM-backed innovations and their growing autonomy promising developments that will require new policies and practices [14]. One has the heavier noun and the other the firmer verb, and an operator setting a compliance budget now has to decide which of the two to plan around.
The obvious reading is that agent-specific rules are coming, and I'd expect them to cover who authorized a payment and who bears the loss when an agent pays wrongly. A second reading rests on his remark that the shift could also be used to "test and refine new approaches" [7]. That sounds like supervised trials before a rulebook. A third reading is that attention arrives through supervision that already exists, of sanctions screening, anti-money-laundering controls and cyber defense, where his remarks were most specific [8][10].
I'd put the most weight on the third. His screening point rested on research, by his account [9], while his agent remarks stayed at the level of a list [6]. On cyber he went further. "Threat actors need to exploit only one key vulnerability, whereas payment system operators and service providers need to defend a large attack surface," he said [11]. "That asymmetry is precisely why the industry must move deliberately to leverage AI in strengthening the safety of our trusted payment systems" [11].
The counter-thesis is institutional. Waller oversees the central bank's payments operations for the Board of Governors [1], and American Banker calls him one of the Fed's longest-tenured policymakers [13]. A change in how often and when payments move [4] hits the system he runs before it reaches any bank's compliance team. If the Fed asks for public comment on agent-initiated payments before it issues anything on AI in screening or cyber defense, my view is wrong.
His routing remark set its own condition. "If presented with the right data and criteria, AI agents have the potential to excel at solving these complex optimization problems, helping to improve the efficiency of crossborder payments," he said [12]. The data and the criteria are the operator's to supply, and that work can start before any rule is written. Waller said: "Meeting this moment requires a proactive approach that balances innovation with the safety, integrity, and stability that underpin trust in payments" [5].
What to watch
- Whether the Fed says how machine-to-machine micropayments would clear and settle through the payments operations Waller oversees.
- Whether other US bank supervisors adopt Waller's 'could necessitate new regulations' language in their own remarks on AI in payments.