Keep the bank in the loop when its data leaves
the vendor for a third-party LLM model.
GenJinni is a governance gate between a vendor's AI stack and the bank whose data it holds — every identity, prompt, dataset, and tool is minted, blessed, and checked on-chain before a single call reaches an external model.
How it works
Sets the policy, reviews the exceptions, and reads a tamper-evident record of every call.
Every identity, prompt, dataset, and tool is a token here — proposed, then approved, or rejected.
Mints the tokens, requests the call, and never gets to the model without clearing the gate.
Holds no registry token of its own. It just answers — after everything upstream has cleared.
Three scenarios, one gate
Pre-approved artifacts clear the gate with zero clicks from the bank.
Manual review or a one-artifact escalation — the bank decides, on-chain, either way.
Anchor a call's evidence on-chain, then watch a database edit get caught by its own hash.