Trust Center
Every item below has an owner and a review date. An expired review comes down rather than going stale.
AI governance note
Model versions ship visibly in the product · predictions are labeled with ranges and a model version · abstention is an honest output, not a failure · our training-data policy is written here, not left to a marketing page’s fine print: your workspace’s own data improves your own recommendations by default, and contribution to shared model improvement is explicit, consented, and de-identified.