Skip to main content
BBritemet
Home Estate About Security FAQ Contact
Sign in Start Home free
Legal

AI Policy

Version 1.2 · Effective: 21 July 2026 · Covers Brit and related AI features in Britemet Home · Privacy Policy · Terms of Service

This policy describes how AI is designed to work in Britemet Home based on current product behavior and code-level safeguards. It is not legal advice. Have counsel review before launch, and treat model providers’ and cloud providers’ terms as additional obligations for AI processing.

1. How we value your privacy

Britemet Home is built so that running a household does not require giving an AI model free rein over your private life. We separate three ideas on purpose:

  • Your system of record stays in Britemet Home under your household’s roles and encryption controls;
  • Brit and related AI features are optional helpers that only call a model when AI is enabled for the household and a member uses a feature that needs it;
  • Highly sensitive vault fields are designed never to be packed into AI context, even when AI is on.

We do not sell personal data for advertising. We do not treat your household archive as a public training corpus for our own general-purpose models. AI is a feature of the product, not the product’s reason for holding your data. Full account and retention rules are in the Privacy Policy.

2. What Brit is (and what else uses AI)

Brit is the optional AI assistant built into Britemet Home. It helps household members ask questions about records the household keeps in Britemet Home, and can help create or update everyday operational items such as reminders, tasks, or related actions when the product allows it.

The same household AI controls also cover other optional model-assisted features when they are turned on, including (where offered) nameplate / capture OCR, inspection-report extraction, inbox document extraction, repair photo identification, and AI-assisted weather emergency checklists. If AI is off for the household, those features stay offline or use local non-model behavior only.

Brit is not a separate consumer AI app. It is part of the authenticated Home product and is intended to answer from your household’s own records, subject to your role permissions, not from unrestricted browsing of the public internet as your system of record.

3. Opting out of AI for the whole household

You can use Britemet Home without any model calls. Core household sections (maintenance, inventory, documents, family roles, emergency card, and the rest of the product surface) are not conditioned on talking to an AI model.

Ways to stay off AI processing:

  • Household opt-out (recommended control). Owners and admins can turn AI off for the entire household in Settings → AI & privacy. When off, Brit runs local search only, and other model-bound features (document extraction, vision OCR, AI weather checklists, and similar) do not send data to Claude. Changes are audited for the household.
  • Simply do not use Ask Brit or AI-consent features. If you never open the assistant or never approve an AI extraction opt-in, we do not build model context for that activity.
  • Prefer non-AI setup paths when offered (for example a standard form instead of AI-guided onboarding).
  • Email request. If you need help applying or confirming opt-out, email privacy@britemet.com or hello@britemet.com.
  • Member choice. Even when AI is enabled for a household, an individual member can decline to use Brit. Role limits still apply if they do use it. Some document or photo features also require a separate per-action confirmation before content is sent for extraction.

Opting out does not delete your household records. It only stops optional AI assistance and related model calls. Account deletion and export remain separate rights under the Privacy Policy.

4. Who can use AI (when enabled)

  • Only signed-in users with access to a household, and only when AI is not disabled for that household;
  • Context sent to the assistant is role-scoped: children, sitters, and other limited roles do not receive finance, documents, inventory, or other areas they cannot see in the product UI;
  • Use may be rate-limited to protect the service and control cost and abuse.

5. Data classes: what may go to a model

We design a hard line between what lives in Britemet Home and what may be sent to a model. The product enforces three classes:

Data class Examples Where it goes
Vault / medical Wi‑Fi passwords, safe combinations, alarm / door codes, medical vault fields Database only. Never included in model context. Enforced in application code when building AI payloads, not by asking the model to ignore secrets.
Operational records Maintenance, vendors, inventory, appliances, service history, and similar non-vault household records (subject to role) May be included in model context only if household AI is on (opted in / not opted out) and the member’s role may see that class in the UI.
Free-text you type Chat messages, free-form descriptions you enter for AI-assisted features May go to the model when AI is on. The product warns you in chat and related flows: do not paste passwords, codes, or medical secrets into free text.

5.1 Operational context when AI is on

When you actively use Brit (and AI is enabled), the service builds a compact, role-filtered snapshot. Depending on your permissions and what you have stored, that may include non-vault operational context such as:

  • Home profile details you entered (for example address-level home information);
  • Maintenance tasks, service history, appliances, vendors;
  • Inventory and document metadata (if your role may view them);
  • Calendar events, chores, todos, projects;
  • Family and pet records, school contacts (if permitted);
  • Vehicles and service history;
  • Travel and sitter guide content (as permitted);
  • Finance-related items such as unpaid bills or subscriptions (only if your role may view finance);
  • The text of your question (prompt).

5.2 Vault fields that never go to the model

Assistant context is built so that it excludes encrypted vault secrets and medical vault contents, including:

  • Wi-Fi passwords and similar stored secrets;
  • Safe combinations and alarm codes;
  • Medical details stored in the encrypted medical vault.

Those tables are not queried when building model context. Vault data remains in your household system of record under role and encryption rules.

5.3 What that means in practice

  • Storing a safe code or medical note in Home does not by itself send that value to Claude;
  • Using Home without Brit (or with AI turned off) does not send household snapshots to Claude for chat;
  • Turning AI off in Settings stops outbound model calls for that household (assistant and other gated AI features);
  • Roles still gate what any user, including Brit on their behalf, can see;
  • Free-text you type is different from vault fields: avoid pasting secrets into chat even when AI is on.

6. How data is transmitted to Claude (AWS Bedrock in production)

We keep Claude as the model family, with stronger account tenancy in production:

  • Production: Claude is invoked through Amazon Bedrock in our AWS account, using the application’s IAM task role (not a shared consumer API key in the browser). That improves tenancy and auditability relative to a single shared third-party API key sitting only outside our cloud account.
  • Local or non-production environments: may call the Anthropic API directly when an API key is configured for development and testing.
  • Offline / not configured / household opted out: Brit and related features degrade to local search or non-model behavior so fewer or no model transmissions occur.

When full AI mode is configured and you use Brit (with AI not opted out), processing works roughly as follows:

  • Who initiates: Your browser talks only to Britemet Home over HTTPS. You do not call Bedrock or Anthropic directly from the marketing site or the browser with your vault keys.
  • Server-side call: Our application servers on AWS assemble the permitted prompt + role-scoped context and send that payload to Claude via Amazon Bedrock (production) over an encrypted connection (TLS), under our AWS account and IAM controls.
  • Purpose: Generate an assistant reply for that request (and, where the product allows, structured help such as drafting a reminder, or consented document/vision extraction). We do not send AI context so that a model provider can market to you.
  • What is not sent: Encrypted vault secret fields and medical vault fields are designed to be omitted from the outbound context (see section 5).
  • Sub-processor role: AWS (Bedrock hosting) and, where the Anthropic direct API is used, Anthropic, act as processors / sub-processors for this limited purpose under our configuration and their terms. Review AWS and Anthropic commercial and privacy terms for provider-side retention, abuse monitoring, and training commitments applicable to the path you use.

International transfer: AWS region choice and related infrastructure may process data in regions outside your home country. See the Privacy Policy on international transfers and safeguards.

7. What Brit is not

  • Not a doctor, lawyer, financial adviser, or emergency service;
  • Not a guarantee of correctness: answers can be incomplete, outdated, or wrong;
  • Not a substitute for reading source records in the product when accuracy matters;
  • Not authorized to change household roles or bypass security controls;
  • Not a requirement for keeping your household data in Britemet Home;
  • Not trained by us as a public chat product on your private household as open training data (provider-side rules for model traffic are governed by AWS Bedrock / Anthropic terms for the path in use).

8. Actions Brit may take

Where the product allows, Brit may help you create or update operational items (for example reminders or tasks) based on your request. You should review automated suggestions before relying on them. You remain responsible for the accuracy of household records and for any real-world actions you take.

9. Human control and permissions

Brit respects the same permission model as the rest of Britemet Home. A user cannot use Brit to access records outside their role. Owners and admins remain responsible for who they invite, what roles they assign, and whether AI stays enabled for the household in Settings → AI & privacy.

10. Data retention for AI

We minimize how long AI-specific data is kept, separate from your long-lived household records:

  • In-flight processing: Prompts, permitted context, and model responses are processed to return a reply for your session. We do not keep a permanent “training set” of your household in Britemet for building a general model.
  • Application logs: Limited operational metadata (for example timestamps, success/failure, approximate usage for rate limits or cost control) may be retained for security, abuse prevention, debugging, and reliability. Retention for operational logs is typically measured in weeks to a few years depending on the log class; security-relevant trails may last longer when required (see Privacy Policy).
  • Provider-side retention: AWS Bedrock and/or Anthropic may process and temporarily retain model traffic under their own policies for abuse monitoring, legal compliance, and service operation. That retention is not the same as your household database retention. Opting out of AI reduces new transmissions to those providers.
  • Household data retention: Maintenance records, documents, vault fields, and other Home data follow the Privacy Policy retention standards (account lifetime, deletion grace period, longer audit retention where required). Those records are not deleted merely because you stop using Brit.
  • After opt-out or account deletion: Disabling AI stops new model calls for that household (once applied). Deleting your account removes primary application data on the schedule in the Privacy Policy; backups and required logs may lag or remain where the law requires.

11. Your responsibilities

  • Review important answers against source records in Home;
  • Do not paste unnecessary sensitive secrets into prompts (chat free-text can reach the model when AI is on);
  • Do not use Brit to harass, commit fraud, or violate the law or our Terms;
  • Supervise children’s use of the household account and assistant features;
  • Understand that AI output is assistive, not authoritative;
  • Use Settings → AI & privacy, or contact us, if you do not want household data processed through AI.

12. Safety and limitations

AI systems can hallucinate facts, invent vendors or dates, or misread context. Britemet is not liable for decisions made solely on AI output, as further limited in the Terms of Service. For emergencies, contact local emergency services, not Brit. Repair and safety triage in the product uses deterministic rules for life-safety guidance; vision models do not decide emergency responses.

13. Britemet Estate

This policy focuses on Britemet Home. Britemet Estate is intended to follow a fuller secure process (including stricter isolation, dual control, audit, and contractual AI terms where appropriate). Those terms and technical controls will be provided with Estate when offered; Home’s household opt-out and data-class rules are the consumer baseline, not the final Estate bar.

14. Changes

We may update this AI Policy as the product and providers evolve. Material changes will be reflected in the effective date and, where appropriate, notified to account holders.

15. Contact

Privacy and AI opt-out requests: privacy@britemet.com · Product questions: hello@britemet.com

Related: Privacy Policy · Terms of Service · Security · Meet Brit

BBritemet

Least privilege by design. Continuity by default.

Products

Britemet Home Britemet Estate

Company

About Security FAQ Contact

Legal

Privacy Policy Terms of Service AI Policy
© 2026 Britemet Software Privacy · Terms · AI Policy