Your buyers are asking what your AI did. Answer with evidence, not a spreadsheet.
Notaree produces tamper-evident, independently verifiable records of what your AI systems and agents did — so a security review or a US AI record rule gets answered with proof instead of a promise.
Both are free, need no account, and never send your answers anywhere.
Every action becomes a link no one can quietly rewrite.
Each record commits to the hash of the record before it. Change anything after the fact and every hash downstream stops matching.
Batches are committed to an RFC 6962 Merkle root and signed in a hardware key store. Anyone you hand a bundle to can check it offline with our open-source verifier — no account, no call to our servers, no need to trust us.
Two deadlines are arriving at the same time.
The deal in security review
Enterprise questionnaires now carry an AI-governance section. “We keep logs” stalls; a retained, verifiable record moves.
US state AI record rules
Colorado’s SB 26-189 takes effect 2027-01-01 with a multi-year record expectation and no small-company exemption. California’s transparency rules phase in alongside it.
Agents that actually act
Once your AI takes real actions rather than drafting text, “what did it do” stops being a debugging question and becomes an evidentiary one.
Informational only. Nothing here determines whether a law applies to you — confirm that with qualified counsel.
Three steps, and the first one is free.
- Tell us your situationYour AI systems, the data they touch, your buyers, your actual questionnaire. You get a plan built for you — not a checklist.
- Publish a Trust Center pageA page you can hand a security reviewer today, stating honestly what you can evidence and what you are still building.
- Install the SDK and make it trueOne line. From then on every AI action is a signed, retained, independently verifiable record.
// one line, zero runtime dependencies import { notaree } from "@notaree/sdk"; await notaree.record({ system: "underwriter", action, outcome });
You never have to trust us.
Every bundle verifies offline against an open specification, with two independent implementations — TypeScript and zero-dependency Python — that reproduce a committed conformance corpus byte for byte. If our own infrastructure were fully compromised, a verifier would still detect any alteration of existing evidence.
We do not hold SOC 2 yet and will not claim it until we do. We store no customer LLM provider keys. Hash-only mode means we need never see your prompts at all.