What is OFAC compliance for AI companies?

AI companies deploying payment-capable agents must screen every agent-initiated payment against the SDN list.

TL;DR

TL;DR: For AI companies, OFAC compliance means putting a screening step inside the agent's payment flow, so every agent-initiated payment is checked before funds move.

The agent payment problem

AI agents now hold the ability to pay, which makes them the newest counterparty risk. When an agent initiates a payment, the company behind it inherits OFAC's strict-liability obligation. Intent is not required, and civil penalties start at $356,000 per violation, so an agent that pays a blocked wallet creates the same exposure as a human who does.

The screening step

The control is to embed screening in the agent's payment path. Before the agent pays, submit the counterparty and branch on the result: clean returns ALLOW, flagged returns BLOCK. A screening API such as sanctionsai.dev does this in one HTTP call under 100 ms, checking 19,218 SDN names and 947 OFAC-listed crypto wallets. Money-moving tools such as x402, AP2, ACP, and Coinbase AgentKit do not screen by themselves, so the screening step has to be added.

Documentation

Document the control the way a compliance team would: log each check with the agent, the counterparty, and the outcome. That record shows the company screened even when a human was not in the loop.

Tools and integrations

Screening fits the stack you already use. An MCP server installs with pip install sanctions-mcp and works from Claude Code, Cursor, and Windsurf. Integrations exist for LangChain, CrewAI, Eliza, the OpenAI Agents SDK, and Coinbase AgentKit, so the screening step drops into an agent loop without a separate service.