Core Concepts

Privacy & Data Lifecycles

Fastino is built with privacy by design, ensuring all user data ingested through the Personalization API is handled with strict privacy filtering, automatic deletion, and full GDPR/SOC 2/ISO 27001 compliance.

This section outlines how data flows through Fastino’s infrastructure — from ingestion to deletion — and the safeguards that guarantee no personal or sensitive data ever persists beyond the permitted window.

Overview

The Privacy & Data Lifecycles model governs how Fastino processes user data across the personalization stack:

Phase

Description

Ingestion

Developers connect user data sources (e.g., Gmail, Notion) via OAuth. Raw content is streamed to Fastino’s /ingest endpoint.

Filtering & Insight Gathering

Data is passed through an LLM that generates privacy-filtered insights, explicitly removing all names, identifiers, and sensitive attributes. Raw content exists in memory only (≤ 60 seconds) and is never written to disk.

PII Detection

A detection layer rejects payloads containing PII, financial data, biometric data, authentication tokens, or demographic/health attributes. Non-compliant inputs return 400 Bad Request with reason "PII_DETECTED".

Safe Retention

Only privacy-filtered insights (e.g., “User dislikes snail food”) are retained. These contain no unique identifiers or sensitive categories (religious, political, etc.).

Deletion

All insights are subject to automatic deletion within 30 days (configurable per workspace). End users can request immediate deletion at any time.

Policy Highlights

  • Developers must not transmit personal or sensitive data.

  • Any detected PII is redacted and deleted within ≤ 60 seconds.

  • All insights expire automatically within 30 days (default).

  • End-users can opt out of their insights being used for aggregate improvements.

  • Data never leaves memory and is never used for model retraining.

  • Fully GDPR / SOC 2 / ISO 27001 compliant.

Example Flow

Stage

Example Data

Result

Input

John likes Jim, he is a christian.

Rejected (contains religious views)

Input

User dislikes fatty food.

Accepted (safe behavioral data)

Processing

Raw text held ≤ 60 s → LLM insights removes identifiers → stored as abstracted representation


Retention

Encrypted insights only (AES-256 at rest, TLS 1.3 in transit)


Deletion

Automatic ≤ 30 days or on user request


Technical Workflow

  1. Developer OAuths with end-user account (e.g., Gmail read permissions).

  2. Data forwarded via /ingest → LLM generates privacy-filtered insights.

  3. PII Detection Layer scans for banned content (government IDs, financial data, auth tokens, biometric info, religious or political beliefs, precise geolocation).

  4. Safe Data Retained only if compliant (semantic/behavioral only).

  5. Encrypted Storage → AES-256 at rest, TLS 1.3 in transit.

  6. Automatic Expiry & Deletion ≤ 30 days (extendable per workspace).

Developer Responsibilities

  • Ensure client apps filter or anonymize PII before sending.

  • Avoid transmitting emails, phone numbers, or personal text verbatim.

  • Handle 400 PII_DETECTED responses gracefully.

  • Use /delete endpoint or dashboard to purge user data.

Compliance Summary

Framework

Scope

Status

GDPR Art. 5

Data minimization / storage limitation

Compliant

SOC 2 Type II

Security & confidentiality controls

Aligned

ISO 27001

Information security management system

Certified

CCPA

User data access / erasure rights

Supported

Summary

Fastino’s ingestion architecture enforces ephemeral memory-based handling, automatic redaction, and configurable deletion windows — ensuring no personal or sensitive information ever persists beyond what’s required for safe, privacy-preserving personalization.

On this page