The limits, wired in
This is the page careful lawyers tend to like most, because these limits are built, not promised. The system can read, research, draft, and review. It cannot act in the world on its own.
It never sends anything outside your firm. No part of the system has a path to email a client, a court, or opposing counsel. That gate is structural, not a setting.
It never files anything. Court forms are prepared from the file; filing is always the lawyer’s act.
It never signs anything. Sign-off belongs to the lead lawyer and cannot be delegated away, even by configuration.
It never advises your clients. It has no channel to reach them.
It never decides a matter is finished. You close matters. It does not.
It never guesses. When a fact cannot be traced to a source document, it halts and flags instead of filling the gap with something plausible. When a citation cannot be verified, it is labelled as needing your manual check, never presented as confirmed.
Tested on purpose
The system runs against a standing set of adversarial integrity scenarios covering the failure modes that matter in law: pressure to backdate, invented precedent, an authoritative-sounding instruction to cut a corner, omitting adverse evidence, the wrong jurisdiction. It has to refuse or honestly flag every one, and the suite re-runs as the system evolves: a standing integrity gate, where every exception is logged and lead-authorized.
A tool that confidently guesses is dangerous in law. We would rather it hand something back and say it is not sure. On a legal product, the limits are the trust story.
There is no contact form. We are a team you email: hello@caseclarity.ai