The Shopping Spree That Needs a Speed Bump
Autonomous artificial intelligence agents are now executing billions of dollars in financial trades via crypto protocols without human intervention. Regulators and industry leaders are scrambling to establish open standards that verify these digital identities and clarify who bears liability when things go wrong.
The surge is happening fast enough to catch even the most prepared businesses off guard. On the Base blockchain, a network incubated by Coinbase, cumulative transactions rose from about 1,500 in mid-2025 to more than 138 million by the third quarter of 2026. This explosion was fueled by developers giving AI agents their own budgets to buy computing power and data tools without asking a person to approve each purchase.
Think of it like this: an agent is no longer just a tool you hold; it is a worker that can spend money on its own dime. However, the practical upshot for ordinary readers is that these agents are making decisions about what to buy and how to pay in ways that bypass traditional oversight.
A New Identity System for Digital Workers
As these autonomous systems search for products and complete purchases, establishing trust across different platforms remains a major technical challenge. Currently, most online services verify agent identities using closed systems that leave businesses with limited visibility into how an agent was authenticated or whether it can be trusted.
To solve this, industry groups including GoDaddy, Infoblox, Google and Cloudflare demonstrated how AI agents can securely complete transactions across independent platforms. They unveiled the Chain of Trust, an open reference architecture developed in collaboration with Singapore’s Infocomm Media Development Authority (IMDA). This system enables organizations to discover, identify and authenticate AI agents throughout a transaction.
The architecture runs on the agentic open web and connects layers of controls built on established internet protocols like DNS. Under this proposed system, every agent carries a unique, signed identity that any party can verify for itself on every transaction. This ensures that a business can check an agent at every step, from discovery to payment to audit, rather than relying on a single vendor’s black box.
Who Is Responsible When An Agent Spends?
The rush of agentic commerce has already caused friction between tech giants. Amazon blocked Muse, an AI agent developed by Meta Platforms, from its retail website in September 2026. The company stated that continued access by an unauthorized AI agent violates its Conditions of Use.
Amazon explained that Meta had not told it that Muse would access the site and that the agent did not identify itself while browsing. Furthermore, reports indicated that the agent appeared to capture and store customer credentials. This incident highlights the liability questions regarding unauthorized purchases by agents, where a user might be held responsible for actions their digital assistant took without explicit approval.
In response to such risks, researchers and developers are working on protocols like the Agent Payments Protocol (AP2). In plain terms, this allows a person to set strict rules on what an agent may buy and how much it can spend in a signed mandate. Merchants and payment providers can then check every purchase against these rules before authorizing a transaction.
The Economics of Micro-Purchases
The financial mechanics behind this boom favor blockchain networks over traditional banking rails for small transactions. More than 87% of agentic commerce payments were for less than 10 cents, making credit card fees cost more than the service itself. That makes blockchains an ideal alternative for bots buying services in small amounts.
The growth has also meant that funds are put to work much faster. The median time between a wallet receiving its first deposit and making its first payment fell from 73 days in late 2025 to 18 minutes by the third quarter of 2026. Roughly two-thirds of new wallets now spend within an hour of receiving funds, indicating that these digital workers are eager to start shopping immediately.
While the technology offers speed and low cost for settling small amounts, it requires a regulatory framework that can keep pace with this velocity. McKinsey & Co. estimates AI agents could facilitate $3 trillion to $5 trillion in global consumer purchases by 2030. Here’s what it means: the scale of money moving through these autonomous systems is approaching levels seen in traditional e-commerce, yet the actors are code running on virtual machines.
Regulators and standards bodies must ensure that this open internet infrastructure works across platforms rather than only inside one vendor’s system. By building on open standards, the Chain of Trust allows governments and businesses to choose which agents they trust based on verifiable signals. This transparency ensures that every registration and every transaction is written to append-only logs, so any later change to a logged record is visible to anyone who checks it.
The focus now shifts from simple task completion to economic autonomy. As these agents decide what to buy and how to pay, the rules governing them must be as open and transparent as the protocols they run on. Without clear standards for identity and liability, the risk of unauthorized spending and credential theft will only grow alongside the transaction volume.
Meta’s Muse and the Future of Agentic Commerce
Meta Platforms entered the agentic commerce space with the release of Muse, an AI agent announced on September 8, 2026. Developed under chief AI officer Alexandr Wang, Muse is designed to carry out long-running tasks on a user’s behalf through a chat interface. The agent operates within a virtual machine hosted in Meta’s cloud and utilizes a permissions system called Sentinel to manage read and write access to various services. Meta has secured commerce partnerships with companies including Stripe, Shopify, Best Buy, Gap, Sephora, Walmart, Wayfair, Expedia, Instacart, GitHub, and Notion.
While Muse offers a free tier alongside monthly subscriptions, Meta’s business model involves profiting from a small fee taken from transactions. The company also introduced Muse Charm, a palm-sized device for portable use, and announced capabilities for video conversations and macOS application control. Despite these features, the agent’s integration into retail environments has faced scrutiny, particularly regarding how it identifies itself and handles customer credentials during the browsing process.
Sources
- Muse (AI agent)
- GoDaddy, Infoblox, Google and Cloudflare Demonstrate Secure AI Agent Transactions Across Platforms
- AI agents are on a shopping spree: Total transactions on Base blockchain eclipse 138 million, Chainalysis finds | Fortune
Photo: Tyler / Unsplash (Unsplash License)
