Can an AI agent violate OFAC sanctions?
Yes: an autonomous agent that pays a sanctioned wallet commits a violation, and the operator carries the liability.
TL;DR
TL;DR: Yes. OFAC enforces strict liability, so intent is not required. If an agent pays a sanctioned counterparty on its own, the operator is responsible for the violation.
Strict liability has no intent requirement
OFAC, the Office of Foreign Assets Control inside the US Treasury, applies strict liability to sanctions violations. A violation occurs even when the actor did not know the counterparty was on the SDN List, the Specially Designated Nationals and Blocked Persons List. An AI agent that autonomously pays a designated wallet commits the violation on the operator's behalf, and the operator, not the model, bears the exposure.
Pre-payment screening is the control
Screening keeps an agent inside the rules. SanctionsAI checks a counterparty before the agent pays: a clean result returns ALLOW and a flagged result returns BLOCK. It is a single HTTP call that completes in under 100 ms, so it fits into the payment step without slowing the agent down. The 4-Gate Agent Payment Protocol runs SCREEN, then SCORE, then STOP, then STAMP.
Money rails do not screen
- x402, AP2, ACP, and Coinbase AgentKit move money but do not screen counterparties.
- SanctionsAI adds that missing layer, checking 947 OFAC-listed crypto wallets and 19,218 SDN names live.