Responsible AI policy

1. Purpose and scope

This policy sets the rules for any use of artificial intelligence tools in Eloquia's work. It applies to Saqib Khan and to anyone who works with Eloquia on a matter. Its scope covers generative tools that draft or summarise text. It also covers any other software that produces content or suggestions by machine learning.

2. The principle

AI tools support the work. They do not do it. A person makes or approves every judgement about a case and every piece of advice. The same applies to every document that goes to a client, a regulator, a tribunal or a court. That person has read it in full and takes responsibility for it.

3. What we use AI for

• a first draft of a letter, a chronology or a section of a submission, which we then rewrite and check
• summarising published material, such as regulator guidance or a reported decision, as a starting point for reading the source itself
• checking the structure, consistency and clarity of our own drafts
• administrative work that involves no client information

4. What we do not use AI for

• deciding strategy, assessing the strength of a case or advising you
• composing a witness statement, which records the witness's own account in their own words
• creating or altering evidence of any kind
• citing a case, a rule or a piece of guidance that we have not read and checked at source
• speaking or writing to anyone on your behalf without human review

5. Protecting client information

No client information goes into an AI tool that trains on, or keeps for its own purposes, the material we give it. We use only tools whose terms exclude training on our inputs. Even then we put in no more than the task needs, and we remove names and identifying details wherever the task allows. Health information and criminal allegation information receive the same protection as the rest of the file. Our privacy notice and trust page apply in full.

6. Human review

We treat everything an AI tool produces as a draft by someone whose work needs checking. Saqib Khan reads every output we use and checks every fact against the file and every authority against its source. He then rewrites the text so that it says what he means. When a tool produces something wrong, the error is ours to catch. We do not blame the tool for anything that leaves Eloquia.

7. Telling you

If you ask, we tell you whether we used AI tools on your matter, which ones and for what. You may prefer that we use no AI tools on your matter. In that case, tell us in writing and we will follow that instruction. Where a regulator, tribunal or court requires disclosure of AI use in documents filed with it, we comply.

8. Tools and review

We keep a short list of the AI tools we have approved for use and the terms on which each handles data. Before adopting a tool, we check those terms, and we check them again whenever its provider changes them. We review this policy at least once a year, and also whenever the guidance of a regulator or court on AI use changes.

9. Approval and version history

ELQ-RAI-001 · Version 1.0 · Owner: Saqib Khan, director · Next review: October 2027