Responsible AI at Atherna

Atherna is built so that its outputs can be checked. Every claim it produces is tied to a source, with paragraph-level citations, so you can verify the basis for an answer rather than take it on trust. Grounding the model in your documents and primary law is how we reduce the risk of fabricated or unsupported output.

Atherna assists legal professionals; it does not replace their judgment. Its outputs are drafts and analysis for a qualified person to review, not final advice. Plan mode and the built-in editor are designed to keep a human in the loop at every step, with the lawyer deciding what to accept, change, or reject.

We do not train our models on your data. Customer documents and matter data are never retained beyond your session and are never used to improve our models or anyone else's. Your work stays isolated to your environment. Our Security page sets out the details.

Confidentiality is the foundation of legal work, so it is the foundation of how we build. Atherna was the first legal AI to clear a bank's information-security review, and it maintains matter-level walls, role-based access, and a complete audit trail so privileged work stays protected and accountable.

We are clear about what Atherna does and does not do. Citations and audit logs make its outputs traceable, so a reviewer can see what was asked, what was retrieved, and what changed. We would rather state a limitation plainly than overstate a capability.

We evaluate the system against real legal work and monitor it for quality and failure modes. As models and use cases change, we test, harden, and improve, and we welcome scrutiny from the security, risk, and compliance teams who hold us to account.

Atherna is a tool for professionals. We expect it to be used by qualified people who review its output, apply their own judgment, and remain responsible for the work and the advice they give. It should not be used to make decisions that significantly affect people without meaningful human review.