No model training on your data
LiveWe do not train models on customer content without explicit opt-in. Default is off.
- ✓Opt-out by default at the organisation level
- ✓Per-org toggle with audit trail
- ✓Sub-processor terms align (Anthropic API does not train on inputs)
- ✓Revocation respected for future runs
What it is.
Your documents are not used to train, fine-tune, or improve any model unless you explicitly opt in. The default for every organisation is off. Even our sub-processors (Anthropic) honour this — their consumer-grade training-data policies don't apply to API customers like us.
If you want to opt in (e.g. to help us tune the Indian-law model in exchange for early access), it's a per-organisation toggle in admin settings with a clear acknowledgement.
Three steps.
End to end.
New organisations are opted out by default. No action needed.
Per-organisation toggle in Admin → AI. Clear acknowledgement required.
Toggle off and content already processed is excluded from future training runs.
What you get.
- ✓Opt-out by default at the organisation level
- ✓Per-org toggle with audit trail
- ✓Sub-processor terms align (Anthropic API does not train on inputs)
- ✓Revocation respected for future runs
Quick answers.
We collect aggregated, content-free telemetry to monitor model health. No customer content leaves to power this — only timings and error categories.
More in AI Controls & Guardrails.
User-pasted text is sanitised before reaching the LLM.
Daily and monthly cost ceilings per organisation. Soft warnings, hard cutoffs.
Stable prompt prefixes are cached. Faster responses, lower bill, passed to you.
200K input / 8K output limits with graceful truncation and a clear notice.