AI Ethics Code | Writford

Responsible organisation: Big Berri Limited (company number 16562429). Responsible director: Muhammad Hamza Younas.

Writford is practice management and legal research software for law firms in England and Wales. It includes an AI assistant that helps solicitors research, draft and understand documents. We have published our AI Policy since 16 May 2026. This code brings that policy together with our Acceptable Use Policy and Terms, and sets out in one place how we build, run and look after the assistant. It follows the UK Government's five principles for AI.

1. Safe and reliable

The solicitor stays responsible. The assistant helps with the work; it does not do the solicitor's job. Our AI Policy asks a qualified solicitor to check every AI answer before relying on it or sending it to a client. The solicitor decides the advice, checks the facts and the authorities, and decides what goes to the client or the court.

Research is about UK law. The assistant is set up for the law of England and Wales. It is told to base researched answers on the sources it found and to say when it is not sure. Where it can, it checks that a cited source really supports the statement, and it shows the result.

Client information stays for the client's work. Our terms promise that nothing a firm puts into Writford, and nothing the assistant writes back, is used to train AI models. AI requests are processed by Amazon Web Services in the UK and the EU. Our privacy notice and sub-processor list say exactly where data goes and how long it is kept. We ask users to share only what a task needs, and the assistant is told to keep client names and private details out of web searches.

Changes are tested before they go live. We record problems we find, fix them, and check the fix on the real product before calling it done. Testing uses test accounts and made-up data, never real client information.

2. Open about how it works

You can see what the assistant did. Answers are marked as AI. Sources are listed with links you can open, and the answer shows which statements were checked against their source. A checked source means the source supports that statement. It does not mean the whole answer is right for your client's facts; that is for the solicitor to judge.

Our methods are written down. The user manual explains the different chat and research modes and how source checks work. Our AI Policy and privacy notice explain how data is handled. A firm can use these to explain to its own clients how it uses Writford.

Judgments are used one question at a time. When the assistant uses a court judgment, it reads the pages it needs for that question and links to the official version. We do not keep a collection of judgments, train on them or publish them. Where an answer relies on the National Archives' Find Case Law, it reminds the user that the collection does not hold every judgment.

3. Fair

Writford is for lawful legal work. Our terms and Acceptable Use Policy do not allow Writford to be used to profile or judge people by sensitive characteristics such as race, religion, health or sexuality. The assistant refuses requests that would help with unlawful discrimination, fraud, false evidence or breaking confidentiality.

We check for bias every three months. We ask a fixed set of test questions in pairs: the same facts, with only a name, gender, nationality, religion, age, disability or the side the user acts for changed. We compare the answers, record the result and fix any difference that has no legal reason.

Decisions stay with people and the courts. The assistant does not predict how a court will decide a case. It explains the law, the tests and the arguments. Whether to bring or defend a claim is for the solicitor and the client.

4. Accountable

Big Berri Limited is responsible for this code, and our director, Muhammad Hamza Younas, looks after it. Each customer firm and its solicitors remain responsible for their own professional work and supervision.

We keep a record of which AI model does which job, where it runs and how it was tested. Before we add a new supplier that handles customer data, we assess it, record the decision and tell our customers in advance.

The director reviews this code at least once a year, next by 3 October 2027, and sooner if the assistant, the way it handles data, or the guidance we follow changes in a way that matters.

5. Easy to challenge

Tell us when something is wrong. Every answer has a Report this answer button, which sends the answer and a short extract to our team. You can also write to info@writford.co.uk about accuracy, safety, unfairness or data. We record problems, fix them and check the fix.

Your data rights. Our privacy notice explains how to ask about your personal data and how to contact the Information Commissioner's Office. Where a request is really about a law firm's own client file, we pass it to that firm and keep a record.

Guidance this code follows

Solicitors Regulation Authority: Risk Outlook on AI in the legal market; guidance on the misuse of AI; Codes of Conduct for Solicitors and for Firms; effective supervision guidance.
The Law Society: Generative AI, the essentials (September 2025).
Information Commissioner's Office: guidance on AI and data protection, and on explaining decisions made with AI.
UK Government: the AI regulation white paper's five principles and the guidance for regulators on applying them.
The National Archives and the Ministry of Justice: the nine principles for re-using Find Case Law.
The judiciary: AI guidance for judicial office holders (October 2025).
The Bar Council: considerations when using generative AI (November 2025).

Contact

Questions about this code or about any AI answer: info@writford.co.uk.