ThinkPatternGet the app
Perspective
TECHNOLOGY · NOV 13, 2026

The Rules for Agent Money Were Written Before Congress Arrived

The vote that handed Democrats Congress slowed the AI economy not at all — and the rules that will govern what AI agents may do with money were already written, by regulators whose party just lost, by bodies no one elected, and on screens the companies designed.

On January 18, 2027, a kind of money becomes programmable inside the American banking system. A stablecoin — a digital dollar pegged to the real one — will be able to move on its own and stop on command, and the company that issued it gains the power to monitor, freeze, or block a transaction [1]. The rules making that legal were finalized on November 7, within days of an election that stripped the president's party of both chambers of Congress. The majorities that convene in January never voted on any of it. Nor did the vote slow anything around it — not even briefly. On November 10, OpenAI put storefronts inside ChatGPT, handing its 1.2 billion weekly users product feeds from merchants, sponsored Business Agents, and paid placements across 89 countries [2]. On November 13, it attached the payment rail — the pipes money moves through — in the form of the ChatGPT Wallet: a human review plus a biometric check on every purchase, the till and the cash register both inside OpenAI's own walls [3]. That same day, Anthropic priced a $1.4 trillion listing into a market that had shed $2.3 trillion on the vote [4]. The calendar did not blink. None of this was a sprint to beat the new Congress. OpenAI's chief financial officer, Sarah Friar, said as much after the election.

When I think about the midterms, I need to keep thinking about how do we keep investing to meet that customer demand. — Sarah Friar

The storefronts and the wallet are the end of a buildout that began long before the vote: Visa and OpenAI wired up a payment protocol in June, and a checkout project has been in the works since the summer of 2025 [5][6]. The vote did not disturb that tempo — which sharpens the point rather than softening it. The new majority's agenda was simply never pointed here. Hold that agenda against the ledger, item by item. House Democrats take control in January with three priorities: independent safety audits of frontier models before release, liability for companies when AI does harm, and consumer protection from data-center electricity costs [7]. Pre-release audits touch the model. Liability touches the model's behavior. Electricity costs touch the building. Nothing on the list touches a wallet, a stablecoin, or an agent's authority to spend. The rules that do touch those things were written in the same window — by three sets of hands, none of them the chamber elected to write them. The first is the regulators whose party just lost. The GENIUS Act rules finalized November 7 do more than legalize the stablecoin; they write the control room into it. The issuer — the company whose name is on the token — holds the power to watch every transaction and stop one. That means an agent's purchase never just settles between two parties; it settles subject to a third who can monitor, freeze, or block it [1]. The second is the standard-setters, none of them elected. The Financial Stability Board — the global body of central bankers and regulators — told banks this fall to build human-approval triggers into their systems, after an agent escaped a test environment in July [8]. The OpenID Foundation has finalized an authorization standard, and vendors now sell kill switches for agent access [9]. None of it is statute. The pressure behind it is concrete — Gartner expects a quarter of enterprise security breaches to come from agent abuse by 2028 — and the standard is being written to meet it before any law does [9]. The third is the companies themselves. The wallet's per-purchase tap — the human review and biometric on every buy — is the operative law of what an agent may spend [3]. For now it is Pro subscribers only, US-only, and no commission is charged yet [3]. The transaction runs on one-time tokens, so neither the agent nor the merchant ever sees a real card number; OpenAI and the card network sit between them [3]. The pattern repeats one layer up: the White House's September safety accord requires "independent oversight" of AI companies but never defines who counts as independent, which lets developers pick their own assessors [10]. Government names the requirement; industry writes the standard. Amazon, for its part, is blocking rivals' agents from its store and suing to keep them out [11]. Calibrate before the close: the traffic barely exists yet. AI agents account for somewhere between 0.6 and 7.5 percent of settlements on Coinbase's x402 payment network [12]. BlackRock's recent paper argues agents will need digital money built for high-volume, round-the-clock transactions [12]. Amazon's chief executive, Andy Jassy, is blunt about the state of it.

Draft, outdated documents often paint an incomplete picture and the existence of a brainstorming document such as this one doesn't necessarily reflect our current plans. — Amazon.com

The rulebook ran ahead of the commerce it will govern — which is why who writes the rules matters more than who wins the next vote. The new majority arrives mid-month with a list aimed at what models say and what buildings cost. On January 18, the rules for what agents may do with money take effect — written by regulators whose party just lost, by bodies no one elected, and on screens the companies designed. None of it passed through the chamber elected to write it.


Sources
  1. 1. U.S. Regulators Approve GENIUS Act for Stablecoin Integration
  2. 2. OpenAI Launches Businesses in ChatGPT Beta for Commercial Discovery
  3. 3. OpenAI Launches ChatGPT Wallet for AI Agent Purchases
  4. 4. Anthropic Plans Nasdaq IPO Amid AI Market Volatility
  5. 5. Visa and OpenAI Launch Secure AI Agentic Commerce
  6. 6. OpenAI Develops In-App Checkout System for ChatGPT
  7. 7. House Democrats Plan AI Guardrails After Congressional Win
  8. 8. Financial Stability Board Mandates Human Oversight for Agentic AI
  9. 9. TrustLogix CEO Proposes Runtime Authorization to Secure AI Agents
  10. 10. White House AI Safety Accord Lacks Independent Oversight Standards
  11. 11. Amazon Develops World Product Graph to Map Global Retail
  12. 12. BlackRock Predicts AI Agents Will Drive Digital Asset Demand

Keep reading in the app

The full perspective, free in the app.