Acceptable Use
If we deploy any AI workflow into your firm's environment, the following acceptable-use principles apply alongside the engagement letter.
Privileged data stays in your stack
Workflows are installed on tools you already own. Privileged client information does not leave your environment by default. Any exceptions are listed in writing in the workflow specification before build.
No training on your data
We do not use your firm's data to train any model that is shared across firms. Where a workflow uses a hosted model, we configure it to opt out of provider-side training and retention to the maximum extent the provider supports.
Human in the loop
AI output is treated as a draft. Every workflow includes a human-review checkpoint and a fallback path for when the model is wrong, both specified before build.
Telemetry and audit
Each workflow ships with usage telemetry and an audit log so the firm can see, at any time, which prompts ran, who ran them, and against which inputs.
Prohibited uses
Workflows we build may not be used to generate impersonation content, to circumvent confidentiality obligations to third parties, or to produce material that violates the rules of any bar council or court.