AI Processing Disclosure

1. The short version

PRAMAAN stores customer matter records and documents in India-hosted systems for the live India service. AI inference may process selected matter content outside India through approved processors.

If a matter requires India-only AI inference, the firm can request an India-resident inference policy from the settings panel. The firm chooses the policy, can change it, and PRAMAAN records policy changes plus per-run provider and region details in the audit log.

2. What may be sent for inference

AI features may send prompts, user instructions, selected source excerpts, matter context, metadata needed to route the request, and generated outputs to approved inference processors. PRAMAAN should send only what is needed for the requested AI workflow.

3. Where inference may run

Approved processors may operate infrastructure outside India, including the United States or other global regions. The firm's selected inference policy controls which approved routes PRAMAAN may use for that firm.

Storage residency, model training, provider retention, and inference geography are separate questions. A no-training or zero-retention route does not mean the model ran inside India.

4. Training and retention posture

PRAMAAN does not use customer matter content to train third-party foundation models. Approved AI inference routes are configured for no third-party model training and disclosed zero or limited retention posture.

Provider retention terms and security or abuse-prevention records can vary by route. The current summary is listed on the Approved AI Processors page.

5. Broker and serving processors

PRAMAAN may use a broker or routing layer to reach an approved serving processor. The broker is not the whole trust boundary. The serving processor, model family, region class, training posture, and retention posture also matter.

6. Internal review

PRAMAAN personnel may review limited prompts, responses, workflow traces, and related diagnostics for support, security, debugging, abuse prevention, legal compliance, and approved quality improvement. Access should be limited to authorized personnel and proportionate to the purpose.

Product-quality work should use de-identified, minimized, or synthetic examples where practical. Customer matter content should not be used to train third-party foundation models.

7. Customer responsibility

The firm is responsible for deciding whether it has authority to submit client material to PRAMAAN and whether AI inference outside India is permitted for a matter. The lawyer remains responsible for reviewing all AI-assisted outputs before use.