AI Supplementary Terms

Last updated: October 31, 2026

These terms supplement the Terms of Service and apply when you use AI features.

1. AI proposes. AI features draft content, suggest relations between ideas, and propose actions, such as publishing a page or replying to a message.
2. You approve. A proposal has no effect until an authorized User approves it.
3. Software executes. Approved actions are carried out by deterministic software, the same code that runs when a person takes the action by hand. The AI model does not execute them.
4. Agent Mandates. You may grant an Agent Mandate that lets AI carry out specific kinds of actions without approval, within the scope, budget and time limit you set. Actions within a Mandate are treated as approved by you. Only people can grant, change or revoke a Mandate.

We keep an audit record of AI proposals, approvals, Mandates and resulting actions in your Tenant, so you can see what was proposed, who approved it and what happened.

As between you and us, you own Output generated for you. Output may be similar to output generated for others. AI can be wrong: review Output before you publish it or rely on it, especially facts, prices, legal statements and claims about people.

We do not use your Customer Data or Output to train AI models, and our AI providers do not use it to train theirs.

AI features use the third-party models listed on the Sub-processors page. You must also follow their usage policies, as reflected in our Acceptable Use Policy.

AI features may use credits and rate limits. We may change which models power a feature, as long as the feature continues to work materially as described.

If you use AI to talk with your End Customers (for example in chat or messaging), you are responsible for telling them when the law requires it.