Book an assessment

Safe-use assessments for AI models · India

Know how safely you can use any AI model, and get the gaps fixed.

Pick any of 3,651 AI models, tell us how you plan to use it, and Lekha scores the risk in your own setup: data protection, data location in India, licence, decisions about people and safety controls. Every shortcoming comes with its fix, and our team can carry the fixes out.

The free assessment opens for pilot users we've invited. Now taking 3 pilot customers for early 2027.

Safe-use assessmentExample
ModelProprietary model via the developer's API
Score23 / 100
Sensitive data sent to an outside providerFix: data-processing terms and a redaction gateway
LekhaHigh
Credit decisions with no human reviewFix: review step, reasons and appeals
LekhaHigh
Data must stay in IndiaFix: use the provider's India region in the contract
YouMedium
Customers see generated answersFix: grounding, output filters, red-team test
LekhaMedium
Verdict: redesign first · 5 of 6 serious findings Lekha can fix

The problem

A model that's safe in one setup can be risky in yours

Same model, different risk

A chatbot model is low risk for drafting internal notes and high risk for credit decisions on sensitive data. Generic model ratings miss that.

Facts are hard to find

Licences change, India hosting differs by channel, and most developers don't publish model size or training compute.

Finding gaps isn't fixing them

Teams get a list of risks but not the controls, contract terms and tests that close them before launch.

How it works

From "can we use this?" to launch-ready

  1. Assess

    Choose the model and answer eight questions about your use, data, deployment and oversight. You get a score and verdict in seconds.

  2. See the gaps

    Every shortcoming is graded and explained, with the fix and who should do it, plus safer alternatives if a different model fits better.

  3. Fix

    Lekha carries out the fixes: verification at the source, data-protection controls, safety filters, human-review design and pre-launch testing.

  4. Stay assured

    Models are re-checked daily against new data and re-verified every 90 days, so you hear when a licence, tier or hosting option changes.

Pricing

Assess for free. Pay for expert review and fixes.

Free

₹0
  • Self-serve assessment of any model
  • Full list of shortcomings and fixes
  • Model Atlas with access tiers

Pro

₹7,500 a month
  • Unlimited saved assessments
  • Branded reports for your committee
  • Alerts when a model's licence or tier changes
Most teams start here

Expert assessment

from ₹50,000 per model and use
  • A Lekha reviewer checks your real setup
  • Model verified to Level 3
  • Signed report for your risk committee

Fix and assure

from ₹2,00,000 per project
  • We implement the fixes found
  • Pre-launch safety and misuse testing
  • Optional ongoing assurance from ₹25,000 a month

Prices exclude GST and are indicative for pilots. Pilot customers keep their pricing for the first year.

Book an assessment

Talk to us about a model you plan to use

Tell us which model and what you want to use it for. We'll reply within two working days to set up a 30-minute call.

We use these details only to contact you about Lekha.

Questions

What teams ask us first

Can a model developer pay for a better score?

No. Scores come from a fixed rule set and published verification checks. Developers can't pay to change a model's score, tier or verification level.

Where does the model data come from?

The Atlas combines published research data on notable AI models, refreshed daily, with Lekha's own checks of licences, hosting and developer disclosures.

Does Lekha see our customer data?

No. An assessment needs only how you plan to use a model, not your data. Fix projects run under a separate agreement.

Does an assessment make us compliant?

It shows the gaps and how to close them, but it isn't legal advice. Compliance depends on your own policies and controls; check requirements with your advisers.