Security and privacy menu

Data

AI models and training

Which models Pondros uses, and why your data is not used to train them.

Which models

Pondros’s reasoning, writing and web research run on Anthropic’s Claude models. It uses a faster model for quick replies in Slack and a more capable one for harder work, and you can choose the model in the web app. A separate provider turns text into search indexes, so Pondros can find the right message or document.

Training

  • Anthropic’s commercial terms say it “may not train models on Customer Content from Services.”
  • We don’t host or train models of our own, and don’t train anything on your data.

Web search

When Pondros researches something on the web, the search runs through Anthropic’s own search tool, and the answer lists the sources it used.

Mistakes

AI can be wrong. Every answer in Slack carries a reminder to double-check anything that matters, and Pondros never closes a task on its own reading of a reply. See How replies are read.