Autonomous AI spending forces banks and wallets into compliance rethink
Fireblocks warns corporate cards and bank transfers cannot police AI agents making thousands of micro-payments, raising fresh AML questions.
Enterprises deploying artificial intelligence agents to make autonomous payments are running into a compliance gap that neither corporate cards nor traditional bank transfers were built to close, according to analysis published by digital asset infrastructure firm Fireblocks. The findings point to a fast-approaching regulatory challenge as AI systems begin executing thousands of small transactions without human sign-off, raising questions over how anti-money laundering and know-your-customer obligations will be enforced when the payer is a machine rather than a person.
Fireblocks’ research describes AI agents already operating inside marketing and engineering departments, launching advertising campaigns that trigger payments to multiple vendors from a single budget, and settling dozens of micro-transactions per task to pay for cloud computing and API access. The firm argues that existing payment rails were never designed for this frequency or granularity of spend, creating friction for enterprises that want to scale automated commerce while staying within regulatory bounds.
Cards and transfers built for humans, not machines
Corporate cards allow enterprises to set overall spending limits but cannot enforce rules such as restricting transactions to approved vendors or imposing expiration windows on a given task, Fireblocks notes. Bank transfers offer more granular controls but typically require manual approval for each transaction, a bottleneck that undermines the automation AI agents are meant to deliver.
That mismatch is pushing enterprises towards programmable wallet layers — infrastructure that embeds smart contract logic directly into the payment process, allowing budget caps, counterparty restrictions and expiry conditions to be set at the wallet level rather than relying on post-transaction oversight. Fireblocks frames this shift as an architectural decision for enterprises, rather than a routine procurement choice, given the compliance stakes involved once agents begin transacting across borders.
Wallet providers move to fill the gap
Several providers have already launched infrastructure aimed at this market. MoonPay introduced its Open Wallet Standard in March 2026, giving AI agents a programmable method to hold and move value, while Cobo followed in April 2026 with its Agentic Wallet, which uses multi-party computation to enforce execution boundaries specific to each task an agent is assigned.
Fireblocks also points to Papaya Global, which already runs programmable wallets to handle contractor payments across 180 countries, as a template that could be adapted for agent-driven transactions at similar scale. The comparison underlines that the underlying technology is not new — what is new is applying it to counterparties that are software rather than people, which sharpens the compliance question for regulators still working out how AML and KYC frameworks apply to autonomous systems.
A narrow window for enterprises
Fireblocks estimates that most enterprise AI rollouts are between 12 and 18 months away from needing robust payment capabilities of this kind, suggesting the infrastructure decisions being made now will shape which companies can deploy agents at scale without regulatory friction. Firms that delay risk having to retrofit compliance controls onto systems already handling high transaction volumes, a scenario likely to draw closer attention from financial supervisors in the UK and EU as agentic commerce moves from pilot projects into production.
For institutional readers, the development reinforces a broader pattern: as AI agents take on financial decision-making, the compliance burden shifts from monitoring human behaviour to auditing code, budgets and permission structures embedded in wallets themselves — a change regulators have yet to fully address.
Read more: Safe’s 130m Q2 transactions underline institutional pivot to smart accounts
