Security at Houser

You should never need to enter a tenant’s Social Security number or financial account number to use Houser — it’s decision-support for professionals, not a system of record. To back that up, every message runs through an automated screen, in your browser and on our servers, that detects Social Security numbers and financial account numbers and blocks them before they are sent to the AI models, asking you to remove them first. When a message looks like it may contain other real tenant details, we warn you and ask you to confirm it’s a hypothetical example.

This screen is automated and best-effort — it targets those specific identifiers and is not a guarantee, and it does not block names, addresses, or case narrative. You remain responsible for what you submit. See Privacy Policy §3 for the precise scope and limits.

All traffic to and from Houser is encrypted in transit (TLS), and your data is encrypted at rest. The application runs on infrastructure from providers that maintain their own SOC 2 Type 2 attestations — Supabase (database, vector store, and file storage) and Vercel (application hosting). The Service is operated from and hosted in the United States.

Sign-in, passwords, sessions, and multi-factor authentication are handled by Clerk; we never store your password. Inside the application, every request is checked against the requesting user’s identity and organization, and database access is governed by row-level security. Access decisions are enforced in our application code on every request, not left to the client.

Organizations on our Teams and Enterprise plans get isolated workspaces. One organization’s documents, conversations, and uploads are never reachable by another organization or by individual users outside it — isolation is enforced both by database row-level security and by access guards in application code, and is covered by our test suite.

All AI calls happen on our servers — never directly from your browser. Your inputs are sent to our AI subprocessors (listed below) only to generate your answer. We do not use your content to train our own AI models. In our Replies feature, structured identifiers are tokenized before any text is sent to a model and re-inserted only in your local result.

Houser’s knowledge base is built from public-domain U.S. federal and state housing materials — the CFR, PIH and Housing Notices, IRS guidance, published QAPs, Admin Plans, ACOPs, and similar sources. Because the regulations themselves are public, you don’t need to upload anything sensitive to get cited, authoritative answers.

You can delete individual conversations, or close your account, at any time. When you close your account we delete or de-identify your personal information — account data and conversation content are removed from active systems within approximately 90 days of closure, and operational/usage logs are retained for about 12 months, subject to backups that age out on a rolling basis. Full detail is in Privacy Policy §7.

We apply per-user rate limiting and automated abuse detection across the application, and we limit bulk extraction of our knowledge base so the corpus can’t be scraped wholesale through the product. AI features that incur cost are metered to prevent abuse. Security-relevant events are logged and retained for review.

We use a small set of vetted service providers to operate Houser. Each processes data only to provide its service to us, under its own terms.

Houser is built on SOC 2 Type 2–certified infrastructure (Supabase, Vercel). Houser’s own SOC 2 examination is on our roadmap. We are not a HIPAA-covered entity, and the Service is not designed to receive protected health information. A Data Processing Addendum (DPA) is available to Enterprise customers as part of the contracting process.

We welcome responsible disclosure. If you believe you’ve found a security issue, please email support@aihousers.com with the details and steps to reproduce, and give us a reasonable window to investigate before any public disclosure. Please don’t access or modify other users’ data, or run automated scanning that degrades the Service, while testing.