Can you prove who approved an AI-assisted filing? A practical test
Four questions that reveal whether your firm's AI provenance would survive a malpractice inquiry or a client audit.
The four questions
First: for any given document that left the firm, can you name the model that touched it? Second: can you name the human who released it? Third: can you produce the prompt and source material it drew on? Fourth: can you demonstrate that none of the first three answers were edited after the fact?
Most firms can answer one or two. Answering all four is what turns an AI program from an efficiency story into a defensible one.
Why hash chaining is the cheap answer
Hash-chaining an audit log costs almost nothing to implement and changes the character of the record: each entry commits to the one before it, so silent edits break the chain and become visible. Add periodic signed roots and a third party can verify a snapshot without ever reading the underlying records.
Put the model to work
Vanguard enforces the oversight tiers described here in the software itself.
- 57 specialist agents
- 3 oversight tiers
- Hash-chained audit trail
- EN / ES parity