NyayVaani reads your documents where they live — on your machine — and answers with citations that click through to the exact clause. If a sentence can’t show its source, it doesn’t pass.
Can Meridian assign the SPA to its new parent without our consent?
No — assignment requires prior written consent. Clause 14.2 of the SPA prohibits either party from assigning rights or obligations without the other party’s prior written consent1. The Disclosure Letter carves out one exception — assignment to a wholly-owned subsidiary — which does not extend to a parent entity2. Consent would therefore be required before any assignment upward in the group.
“Neither party may assign, transfer or otherwise deal with any of its rights or obligations under this Agreement without the prior written consent of the other party.”
“Clause 14.2 shall not restrict an assignment by the Purchaser to a wholly-owned subsidiary of the Purchaser.”
Your matter files are read, indexed, and searched on your own machine. There is no document store to breach, because the store is your disk. The only thing that crosses the wire is the request you approve — sent directly to the model provider you chose, under your own key.
Every sentence NyayVaani drafts carries a pin to its source — document, clause, paragraph. Click it and the span is in front of you, highlighted, ready to be judged. Anchors point at paragraphs and clauses rather than page numbers, so they survive reformatting and hold up inside a marked-up draft. And when an answer can’t be anchored, it says so — flagged as unsourced, never dressed up as fact.
Define the columns once — parties, term, change of control, governing law — and run them across the whole data room in parallel. Each cell is extracted, not summarised, and each cell cites the span it came from. Save the run as a workflow and a junior repeats it on the next matter in one click.
| Document | Change of control | Governing law | Term |
|---|---|---|---|
| SPA — Meridian | Consent required ¶ cl. 14.2 | England & Wales ¶ cl. 21.1 | Completion + 24 mo ¶ cl. 3.4 |
| Shareholders’ Agmt | Tag-along triggered ¶ cl. 9.1 | England & Wales ¶ cl. 18.2 | Perpetual ¶ cl. 2.1 |
| Escrow Agreement | Silent flagged · not found | New York ¶ § 12(a) | 18 months ¶ § 4 |
| Supply Agmt (2019) | Termination right ¶ cl. 15.3(b) | Singapore ¶ cl. 22 | Auto-renews yearly ¶ cl. 4.2 |
Most legal AI is a very long prompt asking a model, politely, to behave. NyayVaani is engineered so that misbehaviour has nowhere to go.
Every action the model takes passes through a typed
interface. Invalid output isn’t discouraged by prompt prose —
it’s unrepresentable. The compiler holds the line
so the model doesn’t have to.
Clause and paragraph numbering is computed by an engine in code, never generated by the model. Cross-references stay consistent through every edit, because arithmetic doesn’t hallucinate.
Citation rules load only when documents are attached; drafting rules only when drafting. Smaller prompts mean fewer places to fail — and cheaper, faster runs on routine work.
For a bounded matter, NyayVaani reads the documents — all of them — instead of guessing which fragments a vector index thinks are relevant. Retrieval is where accuracy leaks; so for matter-sized corpora, we don’t retrieve. We read.
NyayVaani is in early access with a small group of transactional lawyers. If your practice reviews data rooms, negotiates agreements, or drafts against precedent, we’d like your verdict.