OFAC screening for agentic commerce
When an AI agent buys or sells on its own, the screening step must be built into the agent itself.
TL;DR
TL;DR: Agentic commerce, where AI agents buy and sell autonomously, must include sanctions screening in every transaction, because no human is reviewing the counterparty before payment.
Why agents change the risk
An agent acts without human review. If screening is not built into its payment step, the agent will pay a blocked party automatically, and strict liability still applies to the operator. The speed that makes agents useful also removes the human checkpoint.
How to screen an agent
Screen the counterparty before the agent pays. sanctionsai.dev returns clean equals ALLOW and flagged equals BLOCK in one HTTP call under 100 ms, so the agent can stop itself before sending value.
Where it plugs in
- Claude Code, Cursor, and Windsurf.
- LangChain, CrewAI, and Eliza.
- OpenAI Agents SDK and Coinbase AgentKit.
- x402 and MCP, via pip install sanctions-mcp.