The short answer
WordCrafter calls the AI providers' paid APIs, and OpenAI, Anthropic and Google do not train on data submitted through theirs. DeepSeek's API terms are the exception, so if that matters for your content, pick a model from one of the other providers. Your content and your research are yours either way: we never train on them.
If you connect your own API key, requests on that key fall under your own agreement with the provider instead. Check your provider's terms for your account: some free tiers allow the provider to use submitted data.
Which features use models WordCrafter chooses
When you write, you pick the models: the Thinker, Writer and Image models in the Agent tab, the model in the Targeted tab, and your defaults in Agent Studio. See Saved instructions and model defaults.
Some features run in the background on models WordCrafter chooses, and those can include DeepSeek, whatever model you picked for writing. They include:
Learning new memories. After each turn with the agent, what you asked for is reviewed for anything lasting about how you write that is worth remembering. See Agent memory.
Tidying your memory, when no other model is set for it.
SERP research, which reads and classifies the pages that rank for your keyword, and builds page plans.
Import and optimize, which can send parts of the content you import to resolve its structure.
Content briefs, generated from your research.
Reading your website to fill in your business context.
So choosing a non-DeepSeek model for writing does not, on its own, keep all of your content away from DeepSeek. If that matters for your work, ask us through support before you rely on it.
Where to read more
The sub-processors list names every AI provider and service that processes data for WordCrafter, and the country where each one processes it.
The Privacy Policy and the Data Processing Agreement set out how your data is handled.
Security and your data covers where your content is stored and who can see it.