Where AI participates in the work, how that participation is disclosed, and who remains accountable for the result.
No hidden AI. Any material drafted, materially rewritten, or scored by AI carries a visible attribution.
Human accountability. A licensed attorney reviews and signs off before AI output is delivered externally.
Traceability. Every AI action is recorded and included in the sealed audit chain.
Documents generated in-app carry a footer naming the AI assistance, the date, and the reviewing attorney.
AI-drafted client messages show an inline badge and require a pre-send review.
Risk scores, insights, and forecasts carry an AI-generated badge with the model and timestamp.
Automations executed by agents appear on the matter timeline with the agent name, action, and rationale.
Requests are routed through a managed AI gateway. Providers may include the Google Gemini and OpenAI model families. The model used for each action is recorded alongside the action itself.
Only the minimum firm context required for the task is sent. Client personal data in prompts is minimized; free-text bodies are sent only when the feature requires them, such as message drafting. Retrieval uses per-firm indices and cross-firm retrieval is blocked at the database layer.
Firms may disable individual AI features in settings. Compliance-required audit and hash-chain functions cannot be disabled, because they are what makes an approval provable.
Suspected AI errors, hallucinations, or attribution failures should be reported in-app through the observability console, or to the compliance contact named in your engagement.
Every agent action in the sandbox carries its attribution footer, and Tier 3 work halts until a named human authorizes it. Run it yourself in under a minute.
