AURORA
Cryptographic evidence for AI decision records
Start new thread

AURORA - The audit trail your AI decisions don't have yet

by•
Your AI made a decision. Weeks later, someone questions it, and you have nothing to show. So AURORA is evidence custody for AI decisions: seal each declared decision record with a qualified, tamper-evident timestamp the moment it happens, so it can't be backdated or quietly rewritten later. Not an AI judge, not another explainability tool, but a custodian. When the dispute comes, you already have the receipt. eIDAS-qualified timestamping, cryptographic hash-chain, independent verification!

Add a comment

Replies

Best
Maker
📌
Hello producthunt! This is HeeJun, solo founder of AURORA. This came out of a question I kept getting stuck on. So, your AI makes a decision. And then, let's say two months later someone asks "wait, why did it do that?" - and what do you actually have to show them? A log you could've edited. A screenshot. Basically "trust me." Nothing that survives an actual argument. So I gave up on building another explainability tool. Another thing that promises to tell you why the AI decided. There's enough of those already, and honestly most of the time nobody's asking why — they're asking can you prove it happened this way, before things went sideways. That's a different problem. That's custody, not explanation. AURORA just seals the decision the moment it's made. Qualified timestamp, tamper-evident, hash-chained, so it can't be backdated or quietly rewritten later. It doesn't judge your AI or grade it. It keeps the receipt. The boring kind you never think about until you don't have one. Built most of this alone, and the hardest part wasn't the crypto. It was not letting it oversell. Every time I reached for a bigger word — verify, prove, guarantee — I had to walk it back to what's actually true. The product kept getting quieter as it got more honest. Harder to sell that way. I think it's still the right call. Still early. Sealing works, timestamps are real, it does what it says and nothing more. Real question for you: if you run AI in production, what happens today when a decision gets challenged? I want to know if this lands, or if I'm solving something people haven't felt yet. Thanks !

Sealed a test record and the downloadable evidence bundle is genuinely useful, packaging the hash, signature, and timestamp into one file auditors can actually hand to legal. The public verification URL is a nice touch for third parties.

Verified a sample decision record and the evidence bundle download was surprisingly thorough, with the RFC 3161 timestamp and RSA signature all clearly laid out. Really practical if you need to prove a specific AI output was made at a certain time.

The verification URL flow is really clean, love that you can hand someone a link and they can check the hash and timestamp themselves without needing an account. Curious how it handles retroactive corrections when an AI system updates its decision logic later.

Maker

 Great question. AURORA never edits or supersedes a sealed record — that's the whole design. Each record is an immutable snapshot of what was declared at seal time (hash-chained + RFC 3161 timestamp). When the AI's decision logic changes later, you seal a new record; the original stays valid as proof of what was decided then. An auditor sees both, with timestamps proving the order. So, I'd say AURORA don't correct history, but it does prove what existed when.