We built the audit trail we kept being asked to produce.
Jiminy started as a personal frustration. AI agents were making consequential decisions and when anyone asked "how did it reach that conclusion?" there was no good answer. We decided to fix that.
The infrastructure for AI accountability doesn't exist yet in any standardised form. Frameworks reference it. Regulators assume it. Auditors ask for it. But most teams are either hand-crafting something bespoke or simply not doing it. Jiminy is the reusable layer that lets every team answer those questions, without building it themselves.
Watch a short overview
How we think about this problem
Every decision should be traceable and challengeable from the moment it's made. Adding it as an afterthought doesn't work.
We produce tamper-evident evidence, not just a claim. A claim without an audit trail is just a claim.
When audits disagree, we take the more cautious verdict. The cost of a false approval in a regulated context is much higher than a false flag.
Your decision traces are yours. We never use them to train or improve audit models. Full stop.
Small team. Big problem. Long game.
Built products across data, analytics, machine learning and AI in automotive, energy, central government, hospitality, recruitment, and more. Spent years watching AI decisions go unaccountable. Decided to do something about it.
Looking for a founding engineer who cares about both evidence integrity and clean Python APIs. Drop us a line.
Supported by advisors from financial services regulation, clinical AI governance, and enterprise legal compliance.
Join the design partner programme.
We're working with a small number of regulated organisations to shape how Jiminy handles your specific accountability obligations. If that's you, let's talk.