Coinbase, Visa, and OpenAI race to put AI agents on money—who controls the next payments stack?

Coinbase has launched “Coinbase for Agents,” a platform that lets AI assistants connect to users’ Coinbase accounts to trade crypto and, over time, access data and make payments or purchases autonomously. The product is positioned for AI agents such as ChatGPT and Claude, effectively turning conversational systems into semi-autonomous financial operators. In parallel, Visa has embedded its payment network inside ChatGPT, enabling the chatbot to independently shop and complete purchases on a user’s behalf. Separately, OpenAI plans to acquire Ona to strengthen its AI coding assistant, Codex, signaling continued investment in the software layer that can operationalize these agent workflows. Strategically, these moves shift control of financial execution from humans and traditional interfaces toward AI-mediated “rails” and agent accounts. That raises geopolitical and regulatory stakes because cross-border payments, crypto custody, and tokenized settlement are increasingly bundled into software experiences controlled by a small set of global tech firms. Institutional finance is also moving in the same direction: Sygnum argues that banking rails are moving beyond the “stablecoin winner” narrative toward multiple tokenized cash instruments that operate interchangeably on a single platform. The winners are likely to be platforms that can aggregate liquidity, identity, and compliance checks while minimizing friction for autonomous actions; the losers are fragmented payment networks and legacy intermediaries that cannot integrate agent authorization and auditability. Market and economic implications are immediate for crypto exchange infrastructure, tokenized cash products, and payment processing. Coinbase’s agent tooling can increase trading frequency and account activity, which typically benefits exchange volumes and related custody and compliance services, while also raising operational and risk-management costs. Visa’s ChatGPT integration points to incremental demand for card-network processing and merchant enablement, potentially affecting payment authorization flows and fraud-detection spend. Sygnum’s “interchangeable tokenized cash” framing suggests a broader market for tokenized treasury instruments rather than a single dominant stablecoin, which could diversify stablecoin demand across issuers and reduce concentration risk. In instruments terms, the near-term sensitivity is likely to be highest in crypto spot liquidity, stablecoin-related rails, and payment network transaction volumes, with second-order effects on fintech compliance tooling and identity verification providers. What to watch next is whether these agent accounts receive tighter guardrails, clearer audit logs, and standardized authorization flows that regulators can supervise. Key indicators include rollout scope (which regions and account types are supported), limits on autonomous spending/trading, and whether exchanges and payment networks publish measurable controls for fraud, sanctions screening, and key management. For OpenAI and its acquisition of Ona, the trigger is whether Codex capabilities translate into faster integration of agent-to-finance connectors and developer tooling that accelerates deployment. Escalation risk would rise if autonomous payments expand without robust consent and monitoring, while de-escalation would follow if platforms adopt interoperable compliance layers and transparent user controls. Over the next 1–3 quarters, the market will likely price in both adoption momentum and the cost of governance, with volatility concentrated in crypto infrastructure and tokenized-cash narratives.

Geopolitical Implications

• 01

  AI-mediated finance concentrates execution power in a small set of global platforms, raising cross-border regulatory leverage.

• 02

  Autonomous spending/trading will likely trigger tighter rules around sanctions screening, identity, and auditability in software-first payment rails.

• 03

  Interoperable tokenized-cash approaches could reduce dependency on a single issuer/network, reshaping bargaining power across stablecoin and banking ecosystems.

Key Signals

• —Rollout scope and eligibility for agent accounts and autonomous spending limits.
• —Published controls for fraud, sanctions screening, and key management tied to agent actions.
• —Developer ecosystem acceleration around Codex/Ona that speeds agent-to-finance integrations.
• —Regulatory responses or guidance targeting AI-driven payment authorization and custody.