The commitments behind the work.
Not training material.
Our policy is not to use your inputs, sources, prompts, reviewer notes, or delivered copy to train or fine-tune AI models. We must use provider terms and account settings that support this and keep training, feedback-sharing, and data-donation options off.
Using information during your own order is different from adding it to a shared knowledge base. Client-specific material should not enter another client’s context or a reusable training dataset.
Limited processing, not “no other companies.”
AI providers and other necessary services may process limited information to write, host, authenticate, take payment, or deliver messages. Only information needed for their role should be sent. These providers must be documented before the service opens.
We do not propose selling your project content, licensing it as data, sharing it with advertisers, or using it in public examples without separate permission. Limited safety monitoring or lawful retention by a provider is not the same as model training.
Human access with a purpose.
Your copy editor needs the brief to review the work. Production access should be restricted to the team members and approved providers handling your project, with customer accounts kept separate. “Private” does not mean no authorized reviewer can see the information.