Responsible AI
Grounded in the record. Honest about the gaps. A lawyer decides.
The rules DhiTantra’s AI follows, so you know what to rely on and what to check.
How answers are made
Six rules the AI follows.
01
Answer from the record
It answers from judgments, statutes and regulations it retrieved, not from memory or the open web.
02
Pin every proposition
Each line ends in the paragraph, section or circular it rests on.
03
Flag what it cannot match
A citation it cannot match to a retrieved passage is marked UNVERIFIED in place.
04
Show treatment
It shows how later decisions treated an authority before you cite it.
05
Stop for a person
Workflows pause wherever a lawyer must decide; nothing is filed without approval.
06
Keep the trail
Every AI step is logged with its inputs, so work can be reviewed later.
What it will not do
Limits, stated plainly.
Found an answer that broke one of these rules? Tell us and we will look at it.
- Invent a case or citationAnything unmatched is flagged, never passed off as checked.NEVER
- Give advice to individuals as if a lawyerConsumer answers explain; they are not legal advice.NEVER
- Send or file without approvalApproval steps cannot be skipped by an agent.NEVER
- Hide uncertaintyWhere authority conflicts or is unclear, it says so.NEVER
Report a problem with an answer.
Every report is read by the team responsible for answer quality.