OFAC screening for AI agent marketplaces
A marketplace that lets AI agents buy and sell services is the intermediary, so it must screen both the buyer agent and the seller agent.
TL;DR
TL;DR: AI agent marketplaces should screen both the buyer and the seller agent for OFAC sanctions before money or services change hands. The marketplace is the intermediary, so it inherits the screening obligation.
Why marketplaces are on the hook
When autonomous agents transact, the marketplace that matches them, holds escrow, or routes payment is functionally a payment processor. OFAC applies strict liability, meaning intent is not required for a violation. If a marketplace connects a sanctioned counterparty to a transaction, it can be treated as having facilitated it, with civil penalties starting at $356,000 per violation.
What to screen for each side
For each agent, screen the wallet address it will pay from or receive into, plus any known identity fields. sanctionsai.dev's screening API returns a clean result (ALLOW) or a flagged result (BLOCK) in a single HTTP call under 100 ms, checking 947 OFAC-listed crypto wallets and 19,218 SDN names from the US Treasury sdn.csv, synced hourly.
Practical integration
- Screen before a match is committed, not after payment settles.
- Use the sanctions_check tool, then risk_score for borderline matches.
- Keep a per-transaction log so each decision is auditable.
The rule is simple: if an agent can pay or be paid on your marketplace, screen it first.
Even a small marketplace inherits the full obligation: the moment escrow holds funds, the platform is the facilitating party, so screening must sit inside the match flow rather than in a manual review queue.