Last updated: April 2026
AI Policy
Literal's AI features are strictly opt-in. You can use Literal — store documents, verify identity, share with consent, sign agreements — without ever enabling AI. This policy explains what happens when you choose to turn it on.
1. AI Is Opt-In by Default
1.1 Disabled Unless You Enable It
AI features are off by default for every Literal account. No document content is ever sent to an AI provider unless you explicitly enable AI in your account settings and initiate a query. Disabling AI stops all AI processing immediately — future queries will not be sent, and no further content is transmitted to any AI provider.
1.2 Three Access Modes
When AI is enabled, you choose how much of your vault Claude can see. There are three modes: Vault Search (Claude can search and parse through your documents to find relevant content when answering a query), Linked Only (Claude only reads documents you explicitly attach in a chat session — no automatic searching), and Disabled (AI is fully off; no document content is sent anywhere). You can switch between these modes at any time in Settings → AI Features.
2. Your AI Provider — Anthropic (Claude)
Literal's AI feature is powered by Claude, an AI model made by Anthropic, PBC. When you submit an AI query, the text of your query and any document content you include are transmitted to and processed by Anthropic's API. This is a necessary part of how AI features work — the processing happens on Anthropic's infrastructure, not inside Literal's encrypted environment. By enabling and using AI features, you agree to Anthropic's Terms of Service and Privacy Policy, available at anthropic.com/legal/privacy.
3. What Anthropic Receives
When you submit an AI query, Anthropic receives the text of your query and any document content Claude accesses to answer it. In Vault Search mode, this may include excerpts from multiple documents in your vault that are relevant to your query. In Linked Only mode, Anthropic receives only the documents you explicitly attach. In either case, Anthropic does not receive your account credentials, vault metadata, or unrelated documents. Anthropic does not use API input and output to train their models by default — review their policy at anthropic.com/legal/privacy.
4. Document Access Modes
Vault Search: Claude actively searches and parses through your vault when answering a query — finding relevant clauses, dates, or details across multiple documents without you having to specify which file to look in. This is the most powerful mode and results in the broadest document access per query. Linked Only: Claude only reads documents you directly attach in the chat. No vault-wide searching occurs. Claude answers based solely on what you provide. This mode gives you precise control over what is sent to Anthropic in any given session. Disabled: All AI features are off. No document content is transmitted to any AI provider.
5. How to Control AI Access
To configure AI access, navigate to Settings → AI Features. From there you can enable or disable AI entirely, and choose between Vault Search and Linked Only modes. Switching from Vault Search to Linked Only or Disabled takes effect immediately — no further vault-wide searches will occur. Previously submitted queries are subject to Anthropic's data retention policies. To disable AI entirely and stop all document transmission, toggle AI off in Settings.
6. Our Recommendation
AI is a powerful tool for understanding complex documents — but it requires sending content outside Literal's encrypted environment. For the most sensitive documents — passports, financial records, legal agreements, medical files — we recommend reviewing Anthropic's privacy policy before enabling AI on those files. Use AI where the utility outweighs the trade-off of third-party processing. That decision is always yours.
7. Changes to This Policy
We may update this AI Policy as our AI features evolve or as we change AI providers. We will notify you of material changes by email or through the product. The date at the top of this page reflects when it was last updated. Continued use of AI features after changes are posted constitutes acceptance of the updated policy.
7. Changes to This Policy
We may update this AI Policy as our AI features evolve or as we change AI providers. We will notify you of material changes by email or through the product. The date at the top of this page reflects when it was last updated. Continued use of AI features after changes are posted constitutes acceptance of the updated policy.