Artificial intelligence terms

Every output is a suggestion. A person with the right role must review and accept it before it affects a matter, a document, money, or anything sent outside the practice. The assistant cannot approve, sign, lodge, release money, or confirm that a person's identity is verified. A set of acts is reserved to people and is blocked in the software. A professional remains responsible for every decision and every document that goes out under their name.

Text produced by an assistant feature is labelled as proposed or generated in the screen where it appears and recorded in the AI activity log. If you copy it into a document or send it to someone, you are responsible for its content. Where the law or your professional rules require you to tell a client, court or counterparty that AI was used, it is for you to do so.

Answers are built from the documents and legal sources in your practice. Where the answer is not there, the assistant should say so. Quoted passages show the source so that you can check the wording. The legal library contains only the sources your practice has added. We do not copy statutes or judgments from publishers or official sites without permission, and the library does not replace the official text.

Language models can produce fluent text that is incorrect, incomplete or out of date, can miss a document or misread a scan, can invent a citation, and can state something with more confidence than is justified. A clause that reads well may not be enforceable. A checklist may miss a step. You must read what the assistant produces, check it against the primary source and apply your professional judgement before relying on it. We do not accept responsibility for loss arising from reliance on unchecked output.

To produce a response, the relevant text from your matter and your request are sent to an AI model provider. Only what is needed is sent. We do not use the content of your matters, or the outputs, to train any model of ours, and we do not authorise the provider to use them to train its models. Requests are logged with who asked and about which matter. Administrators can pause an agent, set limits and review activity in the AI control centre.

The assistant is a productivity tool for licensed professionals. It is not designed or marketed for any use that the EU AI Act classes as prohibited or high-risk, for example assisting a judicial authority to research and interpret facts and law, or making decisions about a person's access to credit, employment or essential services. You must not use it for those purposes, for biometric categorisation or social scoring, or to make a decision about a person that has legal or similarly significant effects without a person's meaningful review.

Automatic scans of client uploads look for readability and inconsistencies, such as a name that does not match, an expired date or a file used before. They cannot prove that a document is genuine, and a clear result is not a guarantee. Identity and screening results are recorded by people after the checks have run.

Sending text to a service provider so that it can answer a request may have consequences for confidentiality and, in some places, for privilege. It is your responsibility to decide, in line with your professional rules and your clients' instructions, which matters and documents may be used with the assistant, and to tell clients where required. You can switch assistant features off for your practice in the AI control centre.

Practice administrators can switch features and agents off, set limits, pause an agent and see the AI activity log. We may suspend an assistant feature at any time for safety or security reasons.