Last updated: June 2026 · Inkwell Technology Studios
This page explains how IAM.bot protects your data. It is an honest account of what is encrypted, what is not, who can access what, and how to report a vulnerability. We will update this page as the architecture evolves.
Encryption
In transit: All connections use TLS 1.2+. Matrix protocol messages use Matrix's built-in encryption at the transport layer.
At rest: Your memory graph and soul.md are stored encrypted in Neo4j AuraDB (AES-256 at rest).
Note on model access: When you send IAM.bot a message, the AI model (Claude, Gemini, or another) must process your plaintext to generate a response. This means the model inference provider receives your message text in-context during that request. We select providers with strong data processing commitments and we do not use your data to train shared models. If this matters to your use case, email privacy@iam.bot to discuss options.
Access controls
Your memory graph is scoped to your identity — no other user can read it.
IAM.bot staff do not have routine access to conversation content.
Graph-tier consent guards prevent cross-user data access by construction.
Deletion propagates to the graph store within 30 days; backups within 90 days.
Subprocessors
Processor
Purpose
Location
Neo4j AuraDB
Memory graph storage
US / EU
Anthropic
Claude model inference
US
Google Cloud Platform
Infrastructure, hosting
US / EU
Matrix.org protocol
Messaging substrate
Self-hosted on GCP
Vulnerability disclosure
If you discover a security vulnerability in IAM.bot, please email security@iam.bot with a description of the issue. We will acknowledge within 48 hours and aim to resolve critical issues within 14 days. We do not pursue legal action against good-faith researchers.
What we are still building
We are actively working on: formal audit logging, a public status page, SOC 2 readiness, and a published security architecture diagram. We will add these to this page as they ship. We do not claim what we have not built.