SHOTGUNTry to forge a receipt.
Other vendors say “trust our AI.” We hand you the cryptography and dare you. Every claim Shotgun speaks is Ed25519-signed, bound to its knowledge source, and chained so each entry locks in the one before it. Edit a word, reorder a claim, drop one, insert one — the chain breaks and the verifier tells you exactly where.
What a signed receipt looks like
After every call, the buyer gets this — “what our AI said on this call.” It verifies with nothing but the public key. No account. No API. No trusting us.
How to take the challenge
Get a real receipt
Book a demo (or put Shotgun on one of your own calls). You'll receive the receipt JSON and the call's public key by email — the same artifact your buyer's counsel would get.
Tamper with it
Change a price. Reword a claim. Delete the deferral. Swap two entries. Splice in a claim Shotgun never made. Anything.
Run the verifier
One command, open source, offline: verify-receipt receipt.json. It checks every Ed25519 signature and the hash chain. Your edit lights up in red, with the exact entry that broke.
The stakes, plainly: if you can produce a receipt that verifies against our public key but contains a claim Shotgun never made, tell us how — we'll credit you publicly and fix it. Nobody has. That's not marketing confidence — it's how Ed25519 works.
Why this matters to your buyer
FRE 707 and the EU AI Act are making AI statements in commercial contexts a liability question. When your competitor's AI says something wrong on a call, there's no record and no accountability — just a dispute. When Shotgun speaks, there's a signed, source-bound, independently verifiable record. Your buyer's counsel stops being a blocker and becomes the person most impressed on the call.
“That's the first time a vendor's offered proof instead of a promise.”
— the line every rep hears after the receipt lands