Deploying a frontier model in a regulated company raises questions that don't exist for individual use: where does data reside, who can access what, how is usage audited, and what happens when a model is updated? Our view is that answering those questions well and upfront is part of deploying AI responsibly.

We built Claude Enterprise around a set of commitments we think any organization should expect from a frontier AI provider. Customer prompts and responses aren't used to train our models by default. Retention is configurable. Administrators get the controls they need to govern access without standing between employees and the work.