My name is Kunjar Bhaduri — 25-year FSI technology executive, former JPMorgan partner, currently building open-source governance tooling for autonomous AI agents in regulated financial services. I'm writing to propose
finserv-agent-audit as a contribution to the FINOS community.
The project is a reference implementation of four governance patterns that every autonomous agent operating under MiFID II, EU AI Act, or SEC Rule 17a-4 constraints needs: a DEFCON risk-state machine with
deterministic 4-level escalation and hysteresis, a Sovereign Veto kill switch that no agent or orchestrator can override, a tamper-evident hash-chained Audit Chain, and explicit statutory mapping to EU AI Act Articles 9/12/14/15 and MiFID II Article 17. It
has zero runtime dependencies, MIT license, Python 3.12+, 94% test coverage, and runs in under 60 seconds. The repo is live today at
github.com/linus10x/finserv-agent-audit.
I believe this aligns directly with FINOS's mandate around responsible and compliant financial technology. The patterns here are framework-agnostic — the same DEFCON state machine and audit chain logic applies whether a firm
is running LangChain, AutoGen, or a bespoke multi-agent stack. I'd welcome the opportunity to present this to the AI Readiness or Compliant Financial Infrastructure working groups, contribute it to the FINOS landscape, or explore a formal project proposal.
Happy to jump on a call at your convenience.
Thank you very much for your attention and consideration.
Kunjar Bhaduri
Stay informed on the latest updates from FINOS, including the FINOS newsletter and notifications of upcoming FINOS events - see