Animoca's Minds runs live Visa payments as Hong Kong weighs AI agents with wallets
Hong Kong regulators have put agentic commerce on the policy agenda just as Animoca Brands' Minds moves real money on Visa's rails, and the answer to who is liable is still wide open.

Minds, the agentic AI platform built by the Web3 firm Animoca Brands, is piloting live transactions on Visa's network while Hong Kong regulators place agentic commerce on the policy agenda. For executives, the pilot marks the moment autonomous software stops recommending purchases and starts completing them, dragging liability, fraud, and dispute rules built for human cardholders into a rewrite.
The thought experiment just became a transaction. Minds, the agentic AI platform built by the Web3 firm Animoca Brands, is piloting live transactions on Visa's network, as reported by the South China Morning Post. Read that again: an autonomous software agent is not merely recommending a product, it is completing a purchase on the same rails that carry your credit card. That is a different category of thing from a chatbot. Conventional chatbots answer questions about products. Agentic commerce relies on autonomous AI systems that browse, compare options, and execute payments on a user's behalf. The gap between "here are three options" and "I bought option two" is where the entire commercial internet is about to be renegotiated.
The timing is not a coincidence. Hong Kong has been pushing to turn automated agents into everyday shopping tools, and local regulators have now placed agentic commerce on the policy agenda. That combination, a live commercial pilot running alongside an active policy conversation, is rarer than it sounds. Historically, payments innovation has arrived first and regulation has arrived later, usually after something breaks. Here the two are moving in the same window, which gives Hong Kong a shot at writing rules that shape how agentic commerce works elsewhere rather than reacting to someone else's standard.
Why Visa's network is the detail that matters most. Card rails were engineered for a human at a checkout: a person taps, clicks, or signs, and every fraud model, dispute process, and liability rule downstream assumes that person exists. An agent that browses, compares, and pays removes the human from the moment of authorization. That single deletion ripples through everything. Fraud detection typically looks for human behavioral signals. Chargeback rules typically assume a cardholder who can testify about what they intended. Merchant risk models typically price the probability that a buyer is who they claim to be. When the buyer is software acting on instructions, each of those assumptions needs a rewrite.
For Animoca Brands, this is less a pivot than a continuation. The company's roots are in Web3, a world built on the premise that digital ownership and programmable money belong in the same stack. An AI agent that needs to hold value, pay for things, and prove it acted on someone's behalf is a natural extension of that thesis. The Web3 framing also explains the wallet metaphor at the center of the story: agents do not just need intelligence, they need a place to keep funds and a way to move them.
The hard questions are the ones regulators will have to answer, and they are not small. Who is liable when an agent buys the wrong size, the wrong quantity, or the wrong thing entirely? How does a merchant verify that the agent actually had permission to spend? What does a chargeback look like when the cardholder is a script? What do know-your-customer and anti-money-laundering checks mean when the account holder is a model acting for a person? These are the questions typically at the center of agentic commerce policy debates, and none of them have settled answers yet. Putting the topic on the policy agenda is the first step toward having any.
Then there are the second-order effects that will hit businesses that never thought of themselves as payments companies. If agents do the browsing and comparing, product discovery shifts from human eyeballs to machine-readable data. Merchants start optimizing listings for software rather than shoppers. Advertising models built on capturing human attention get repriced when the attention belongs to an agent that does not get distracted. Checkout stops being a design problem and becomes an API problem. Every layer of the funnel that assumed a person would pause, consider, and click is now up for renegotiation.
The strategic stakes for executives are straightforward, even if the answers are not. Agentic commerce is the point where AI stops being a demo and starts touching the P&L, because a transaction is the one AI output that is unambiguously measurable. Whoever controls the wallet, and the rules around it, controls a meaningful slice of how commerce gets routed. Boards should be asking three questions now: who authorizes an agent to spend, who carries liability when it spends wrong, and what the audit trail looks like when a regulator or a customer asks. Companies that wait for the answers to be handed to them will be implementing someone else's standard.
Hong Kong's bet is that regulatory clarity is a competitive advantage, not a cost. A live pilot on Visa's rails plus a policy agenda is a small experiment with an outsized precedent: it tests whether a city can move fast enough to shape a category before it hardens. The pilot is small. The precedent is not.
This story's Key Insights and Take-aways are locked.
Create a free account to unlock Executive Actions for one credit.
Register to UnlockAlways free for Executives Club members. Join the Club
More in Technology
Anthropic mathematician claims Claude helped crack the Riemann hypothesis
If the proof holds, AI has toppled math's most famous open problem; if it does not, it is a case study in announcing breakthroughs before anyone verifies them.
AI's $4 Trillion Wipeout Erases Nasdaq's Gains for the Year
A brutal week took roughly $4 trillion off the AI sector and left the Nasdaq flat for the year, and the damage ran the full length of the supply chain.
Anthropic mathematician says AI proved the Riemann hypothesis. Verification will take time.
A single unverified claim from inside Anthropic now sits against more than 160 years of open mathematics, and the gap between asserting a proof and surviving peer review is the entire story.



