Last updated 2026-09-14.

AI use & data collection

How the AI features in RaterOS work, what they see, and the line they never cross.

Which AI we use

Guidance and analysis chat is powered by Claude, made by Anthropic, through Anthropic’s commercial API.

Your data does not train the model

Under Anthropic’s commercial terms, what we send is not used to train their models. Anthropic may keep inputs and outputs for a limited time for safety and abuse monitoring, under its own retention policy.

AI explains. It never sets price.

Rates come from filed rates and written underwriting rules. The AI can explain a result, point to a missing document, or summarise a loss run — but its output is guidance only.

A person makes every pricing, underwriting and appointment decision. No declination, rate increase or other adverse decision is made automatically by AI.

What the AI sees

Only what the task needs: the submission, class and payroll figures, loss history and documents on the record you are working, plus your question. Conversations may be saved to that record so the team can see what guidance was given.

Please don't over-share

Don’t paste medical records, diagnoses, Social Security numbers or other sensitive health or identity details into the chat unless the question truly needs them. A claim amount and injury type is almost always enough.

AI can be wrong

Check anything important against the source document. If an answer looks off, tell us — we fix the rules rather than let the model guess.

More

See Privacy for retention and your rights.

AI use & data · WorkComp RaterOS