Giving an AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → money to spend leaves a separate question: which counterparties may receive it? The payment-screening example described in this talk sends two agents to a paid service, one using an ordinary wallet and the other an address flagged as sanctioned.
According to the description, both payments authorize with screening disabled. Enabling screening rejects the flagged address while allowing the other transaction. The example illustrates a control at the payment boundary: the agent's ability to purchase a service can be constrained by a separate counterparty check. It is a presenters' demonstration, not evidence of comprehensive sanctions compliance.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
agent skill — A reusable instruction file that teaches an agent how to do one job well — the procedure, the tools, and what counts as done.
videoEvals for High-Stakes AI Agents — Doneyli De Jesus & Annabell Schäfer, Langfuse
videoAn AI Agent Got Its User Banned From Resy — Keycard, JPMorgan Chase & AAuth
videoTeach Your Agent to Prove It's Done — Nick Nisi & Zack Proser, WorkOS
videoHow to Know If Your AI Agent Is Any Good: Evals With Phoenix & Claude — Laurie Voss, Arize AI
