Fieldguide security lead's 4 controls for auditable financial agents
AI Engineer · youtube · 2026-11-08
Fieldguide lead security engineer Anirudh Veeraragavan argues that in finance a correct agent answer isn't enough — you must be able to show how it got there. Using a bank reconciliation audit agent, he lays out four controls: authorization preventing agents from overwriting human-reviewed work; execution records enabling step-by-step replay; citations bound to exact document versions; and auditing eval data so a 97% score doesn't hide real failures. He also covers when agents should stop and escalate to humans, and how the pattern extends to credit and anti-money-laundering.
More from coding & agent
- Anthropic's guide: turn Claude into a work assistant with Skills and MCP — Arindam_1729 · 2026-11-08
- Dev benchmarks tiny LLMs on agent tasks: 9B passes 2/5, stepping up to 12B — draginol · 2026-11-08
- Engineer proposes CI/CD could become a handoff to a deployment agent — andreisavu · 2026-11-08
- Dev launches Local AI benchmark scoring models by hardware on agent tasks and speed — draginol · 2026-11-08
- Personal AI dot can now read inbox sign-in codes to log in on your behalf — Dimillian · 2026-11-08
- Dev plans eye-tracker to automatically feed agents visual context — johnlindquist · 2026-11-08