Security and trust
Anthouse is an early company. This page is written so your security team can check it rather than take our word for it, which means the second half is as important as the first. If something is missing here, ask us at security@anthouse.io and we will answer plainly.
Apps you deploy, and the databases behind them, run in the European Union. The workspace record, sharing rules and audit log are in an EU region. Sign-in is the exception and is described below.
We will give you 30 days notice before adding a processor that handles customer data, and this list is the canonical one.
Every secret is encrypted with AES-256-GCM under a per-app data key, which is itself wrapped by a key encryption key that never touches the database. In production that key lives in AWS KMS.
Anthouse reads the source of each version and checks for credentials left in the code, code built from strings at runtime, database queries assembled by concatenation, writes with no identity check, and calls to addresses not on your allow list. Findings name the file and line, and secrets are masked before they are shown back to you.
Middleware and edge gateways are not security boundaries. Every mutating endpoint independently re-verifies the caller against the workspace roster.
An AI agent operating an app inherits that person's role and can never do more. What agents may do per app is set by the app owner or a workspace admin, and agent actions are written to a separate audit trail from human actions.
Ownership is derived from the workspace roster, so when someone leaves, their apps show as unowned rather than quietly continuing to run.
Apps drop to free limits and become read-only. You can export the source at any time.
Listed so you are not surprised by it in a questionnaire.
Not started. We will begin the readiness period when a customer contract requires it, and we will say so here when we do.
Not held.
Your app data and your workspace records are in the EU. Sign-in is currently handled by Clerk in the United States under the EU-US Data Privacy Framework and Standard Contractual Clauses. Moving authentication to an EU-hosted provider is on our roadmap and we will not claim it until it is done.
Not yet commissioned. Planned before the first regulated customer goes live.
Agent permissions are set, recorded and audited today. Refusing a disallowed agent action at the request path itself is in progress. Until it ships, treat the setting as a policy record rather than a hard control.
Coming. Until then, incidents are communicated by email to workspace admins.
Finding apps is the start, not the product. Actions routed through Anthouse are checked against each app's policy at the moment they happen.
Reads pass under a read-only policy. Risky writes wait for a named human to approve — approvals expire in 30 minutes and work exactly once. A revoked app stops answering everywhere, instantly, and can be restored just as fast.
Every decision — allowed or refused — is written to a hash-chained log. Editing or deleting any past entry breaks the chain detectably. Your auditor can verify it without trusting us.
Enforcement covers actions that pass through Anthouse: the MCP endpoint agents use and the data paths apps adopt. Apps we only discovered but that run elsewhere get ownership, posture scoring and evidence — not runtime control — until they migrate onto the pipe. We say this here so you never find it out in an incident.
You should be able to leave. A governance vendor that holds your apps hostage cannot be trusted to govern them.
Anything deployed through Anthouse downloads as a zip of ordinary web files — no proprietary framework, no runtime lock. They run on any static host the moment they are unzipped. This is promised in our terms, and the export endpoint exists today.
Every connector is read-only and revocable on your side — Google admin, GitHub settings, your Entra tenant — in one click, without asking us.
Each app carries a handover pack: owner, version history, who deployed what and when. Registers and event logs export as files. What cannot move is the live, continuously verified evidence chain — history can be copied, not re-created, which is the honest reason customers stay.
If you believe you have found a vulnerability, email security@anthouse.io. We will acknowledge within one business day and keep you updated until it is closed. We will not pursue legal action against good-faith research that does not access other customers' data or degrade the service.
For a DPA, a subprocessor change notification, or a completed security questionnaire, the same address reaches us.