What the AI sees.
When you ask Triplewave to draft or revise, it sends the relevant instructions, document text, facts, brand voice and company info to Anthropic’s commercial API. Neither Triplewave nor, under its commercial terms, Anthropic trains models on that content.
Updated
What goes to the model
Triplewave sends only the context a request needs: the instruction, the document, its facts, your brand voice and your company info.
- Leave secrets and sensitive personal data out of your prompts.
- Read what the model writes, and check every fact that matters.
- You’re responsible for the final document and who receives it.
Chat connectors are separate
If you use Triplewave from ChatGPT, Claude or another MCP client, that app processes the conversation under its own settings and terms. Triplewave limits what the connector can read or change. It can’t change how the chat app handles your data.
- Grant the narrowest permissions that do the job.
- Triplewave remembers something only when you ask it to.
Worth knowing
- The model can be wrong, incomplete or biased.
- We don’t claim zero data retention across every provider, plan or support request.