Attest: an ADK agent that audits what other models say about SEC-registered advisers
A machine is describing your business to prospects several times a day, and there is no record of any of it. That was the whole starting point. I built Attest for the All Things Agentic Hackathon , Fortified Enterprise Fleet track, and I wrote this post for that hackathon too. Code is at github.com/jpka/attest . The problem is narrow on purpose. The SEC's 2026 examination priorities name AI in…
An AI-powered tool named Attest was developed for the All Things Agentic Hackathon to audit how other models describe SEC-registered advisers. The primary goal was to ensure the fairness and accuracy of claims made by firms about their AI usage. Attest performs a regular evaluation of these claims, comparing them against the firm's Form ADV Part 1A filing.
The results are stored in a hash-chained archive, which is append-only and cannot be altered. The agent, built using the Google Agent Development Kit (ADK), runs on Cloud Run and fetches the scheduled evaluations through Cloud Scheduler. It utilizes six tools to perform various tasks, including reading ground truth, scoring answers, appending evidence, and handling per-firm memory.
The scorer model is an older version (gemini-3.1-flash-lite) to maintain consistency with the evaluations performed on the deployed scorer. The agent is designed to be transparent, with every tool call tracked and recorded using Cloud Trace. The system is built to ensure data integrity and prevent overwriting of entries, making it a valuable tool for regulatory compliance and auditing purposes.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
