OFAC Openai Compliance Guide Part 1
OFAC sanctions screening for AI agents built on the OpenAI Agents SDK, covering the basics before an agent pays anyone.
TL;DR
TL;DR: The OpenAI Agents SDK can trigger payments through tool calls but does not screen counterparties against OFAC's SDN List. Add a sanctions check before each payment so clean counterparties ALLOW and flagged ones BLOCK.
Why OpenAI agents need OFAC screening
An agent built on the OpenAI Agents SDK can call tools that move money. OFAC is the Office of Foreign Assets Control, part of the US Treasury. Its SDN List names Specially Designated Nationals and Blocked Persons. Sending value to a listed party is a strict liability offense: intent is not required, and civil penalties start at $356,000 per violation.
The 4-Gate Agent Payment Protocol
Screen every counterparty before the agent pays. The protocol runs SCREEN, SCORE, STOP, and STAMP. A clean result lets the payment ALLOW; a flagged result BLOCKS it. One HTTP call returns a decision in under 100 ms, fast enough for an agent that pays mid-task.
Practical setup
- Wrap the payment tool call with a sanctions_check before transfer.
- Use risk_score to bucket counterparties that need review.
- Start on the free tier, five checks per day, no signup or key.