Security
One cloud, one region
Every byte of your data is processed inside AWS us-east-1: Amazon Transcribe for speech-to-text, Amazon Nova on Bedrock for cleanup, DynamoDB for your settings and history. No third-party subprocessors touch your voice or text.
Audio is never stored
Audio streams from your browser to Amazon Transcribe over TLS (a presigned, short-lived WebSocket) and is transcribed in-flight. Neither SetuInk nor AWS retains the audio. AWS credentials never reach your browser.
Transcripts expire
Dictation history is retained for 30 days (DynamoDB TTL), capped at your last 50 dictations. “Delete all my data” removes everything — history, dictionary, snippets, and usage records — immediately.
Model training
We never train models on your data. Amazon Bedrock, which performs the cleanup step, does not use customer content for training. For transcription we have applied the AWS AI-services opt-out policy so Amazon Transcribe does not retain or use your audio for service improvement. We don't log transcripts outside your history.
Compliance
SetuInk is early-stage and does not yet hold SOC 2 or ISO 27001 certification. What we offer today: a single-cloud, single-region architecture, 30-day retention with one-click deletion, and a data-processing addendum on request for pilot customers. No third-party subprocessor touches your voice or text — with one documented exception: eight regional Indian languages use your browser's own speech recognition rather than ours, which we flag in the language picker and describe in the privacy policy. Formal certification is on the roadmap as the company grows.
Responsible disclosure
Found a vulnerability? A dedicated security contact address ships with our domain launch; disclosure reports will be acknowledged within 48 hours.