Benefit Statement
Where we want to go and how we want to get there
Goals
We founded jhana with a few beliefs that remain core to our operations.
- Access to justice and the quality of justice are highly variable.
- Technological progress holds the potential to expand access to justice while improving the quality and consistency of justice.
- Legal data and technology are highly sensitive assets with broad implications for the public good, and must be handled with care.
- For‑profit entities are best‑suited to pursue technological progress.
We take our beliefs and values seriously, and are doing our best to embed them into jhana’s DNA. jhana, Inc. is a Delaware Public Benefit Corporation (PBC) and jhana Labs Private Limited is a limited company registered in Chennai. Our parent company has registered a set of Stated Benefits. As a PBC, our directors must balance (1) stockholders’ financial interests, (2) the best interests of those materially affected by our conduct—like our users, and (3) the specific public benefits set out in our charter, outlined below.
jhana, Inc.—Benefit Statement
Extract from the biennial statement provided to stockholders as of July 13, 2022:
The Corporation will promote public benefits by:- Democratizing access to information and processes needed by legal professionals and their clients, students, scholars, and other relevant entities around the world;
- Promoting, including via research, open source science, and applications, the advancement of the ethicality, alignment, safety, and benchmarking of artificial intelligence systems;
- Interrogating, reducing, and publicly reporting bias in any models built and/or deployed by the Corporation;
- Protecting the integrity and privacy of training data utilized by the Corporation.
Philosophy & Alignment
Principals and Agents. Principals—people, organizations, governments—seek to secure their rights and intents. Legal systems arose to enable this will. Over time, fraud and abuse pushed us toward dense procedure, and the law has too often become an end in itself. Lawyers act as Agents for Principals, but rote work and needles‑in‑haystacks consume time that should go to strategy and judgment. We aim to denoise the parts of legal work technology can help with, so counsel can focus on decisions that matter for their Principals.
Where we focus first.
Retrieval. Find the right caselaw, pull the clause in diligence, surface the signal fast.
Routine negotiation. Speed the first drafts and redlines so parties can get to what really matters.
Insider knowledge. Make institutional memory—like office histories or opposing‑counsel patterns—available to everyone.
Why legal tech is bottlenecked.
Legal work is process‑dependent; you can’t reverse‑engineer it from outcomes.
Unlike code, there are no unit tests; unlike math, lemmas aren’t provable—supervised training is hard.
Legal knowledge is tacit and tribal; the norms are institutional and learned on‑site.
General‑purpose LLMs are too atomic for the law; multi‑agent coordination with boutique tooling works better.
Much legal data is niche, proprietary, or confidential; annotation at expert level is expensive.
Our approach.
Pre‑training tokens for Indic corpuses and subcontinental knowledge are still primitive vis a vis the west or Mandarin. We digitize, index, and record data that is missing in the common crawl and other common internet corpuses.
We build a stock of data from a history of user interactions, feedback, and edits. This bridges the gap for the missing GitHub in the legal corpus, which makes legal agents vastly underperform at legal tasks vis a vis software editing agents. Our data assets are multi‑lingual, domain‑specific, and emits of annotations by experts.
We build sites for flow of reinforcement learning data, that is incentive‑compatible and in the real world. We capture modalities of input and output and interact with users with real stakes at large N scale.
We observe workflows on‑site and train agent frameworks on interaction data.
We build real‑world tooling for AI agents to access filing systems, data sources, and procedural access points.
Why start in India.
The largest English‑language common‑law market with unmatched volume and diversity of matters.
High potential for platformization where systems of record are still open to be defined.
A pragmatic regulatory posture toward AI and data that welcomes responsible innovation at scale.
Our Work with Courts
A. Neutrality
We do not build outcome‑prediction systems for adjudication. “Who will win?” tools encourage overreach, amplify dataset biases, and risk pressuring courts toward imagined certainties. Generic legal AIs that aren’t bias‑aware can implicitly take sides. Court‑facing AI must be non‑adversarial by design: guardrails for language, strict structure for reportage, and verifiable accuracy.
Our standard: audit trails, not vibes. Every assertion should carry line‑level references, with intra‑document cross‑citation that lets a reader trace claims back to sources and reasoning steps. If it can’t be audited, it doesn’t belong in the record.
B. Continuity of Care
Borrowing from public health, justice should practice continuity of care—systems for the people, not the people for the systems. Concretely, for litigants and legal persons this means:Relationship: a continuous, accountable record that “remembers” the matter’s history.
Mutuality: two‑way outreach for check‑ins, exceptions, and frontline assistance when needed.
Timeliness: access to filings, status, and hearings where and when required (including remote access).
Choice: the ability to explore, choose, and un‑choose representation or pathways without losing context.
Knowledge: plain‑language explanations with layered, fractal detail—so parties stay aware of options, tradeoffs, and progress.
C. Deployable, Composable, Court‑Owned
Court technology should be deployable on‑site, owned by the courts, and built from composable AI blocks—APIs and interfaces that integrate with CMS, e‑filing, and archival systems. Our courtroom stack is designed for modularity and local control. Learn more: https://jhana.ai/courtroom/.Copyright © 2025 jhana.ai. All rights reserved.
Backed By
As seen in
