AI Policy

CleverContracts offers three AI-assisted features, all of which operate on engagement content you explicitly select:

• Summarize — generates a draft recap of selected messages, organized into what happened, explicit decisions, open items, and clarifying questions
• Draft reply — suggests reply text you can review, edit, and then send manually
• Extract change draft — helps turn discussion into a clear, attributable change proposal you can refine before requesting approval

• It does not provide legal advice or enforceability judgments
• It does not auto-apply edits to engagement records — every output requires your explicit review and approval
• It does not auto-send messages to clients
• It does not appear in client-facing screens — clients never see AI features or know they are being used
• CleverContracts does notoffer a “train on my data” feature — your data is never used to train any AI model

AI features are powered by one of the following third-party providers:

• xAI (Grok) — we set the store: falseparameter on every request. Per xAI's enterprise terms, inputs and outputs are not used for model training and are deleted within 30 days.
• Google Vertex AI (Gemini)— per Google Cloud's standard data processing terms, customer data submitted to Vertex AI is not used for model training.

Both providers are located in the United States. Content is transmitted securely via HTTPS/TLS and is processed ephemerally.

AI actions are designed to send the minimum data necessary:

• Each message is truncated to a maximum of 4,000 characters
• A maximum of 50 messages are included per request
• Only the messages you explicitly select are sent — not the entire engagement history
• Attachment contents are not sent to AI providers — only text content is used

AI actions use only the messages you explicitly select. The UI displays a “Sources used” list so you can verify what the output was based on.

The system prompt that guides AI behavior is fixed and designed to produce structured, factual outputs. It does not contain your personal data or any content from other users.

AI outputs are drafts. Review them carefully and edit as needed. Nothing changes until you apply it.

All AI outputs are clearly labeled as “AI-generated” and include a “Review before using” notice. AI outputs are not legal advice.

If AI is unavailable (provider outages, rate limits, invalid output), the platform shows an explicit error state and you can continue manually. No fabricated “fallback” output is ever shown.

AI is optional and off by default. To enable AI:

• You must be on a plan that includes AI features (Pro or during a trial)
• You must explicitly enable AI in your workspace settings, which requires a two-factor authentication step-up
• You must acknowledge the AI policy during enablement

Enabling AI is revocable — you can turn it off at any time in your settings.

AI usage is rate-limited and quota-limited to prevent runaway costs and abuse. If you hit a limit, you will see a clear message and can continue using the platform manually.

Separately from enabling AI, you may opt in to sharing AI diagnostic data. This helps us monitor AI feature quality and performance.

Diagnostic data is strictly content-free — it includes only:

• Which AI action was used (summarize, draft, extract)
• Which AI model was used
• Response latency (how long the AI took)
• Token count (input and output size)

No message content, scope text, or any engagement data is included in diagnostic records. Diagnostic events expire automatically after 30 days.

You can opt out of diagnostic data collection at any time. Existing diagnostic data expires naturally within 30 days.