Can Lawyers Use AI Ethically? What Bar Regulators and Courts Actually Say

Every few weeks a lawyer asks me some version of the same question: am I allowed to use this? The answer, in every major jurisdiction I follow, is yes, with conditions. Regulators have not written a new code for artificial intelligence. They have explained how the duties lawyers already owe apply to a tool that drafts fluently and is sometimes wrong.

This guide summarises what the leading regulators and courts have said, then distils the duties that apply wherever you practise.

The short answer

Generative AI is permitted. What is regulated is how you use it. The failures that have led to sanctions and referrals are failures of existing duties: filing unverified work, exposing confidential information, failing to supervise, or misleading a court. If you keep those duties, you can use AI with confidence.

United States: ABA Formal Opinion 512

On July 29, 2024, the American Bar Association’s Standing Committee on Ethics and Professional Responsibility issued Formal Opinion 512, its first formal opinion on generative AI. It is advisory rather than binding, but state bars and courts draw heavily on it, and many states have issued their own guidance along similar lines.

The opinion works through the Model Rules:

• Competence (Rule 1.1). Lawyers need a reasonable understanding of the capabilities and limitations of the tools they use, and must not rely on output without an appropriate degree of independent verification.
• Confidentiality (Rule 1.6). Before inputting information relating to a representation, a lawyer must evaluate the risk of disclosure. For many self-learning tools, informed client consent may be required, and boilerplate consent in engagement letters is not sufficient for that purpose.
• Communication (Rule 1.4). In some circumstances the client must be told about AI use, for example where it affects the basis of fees or where the client asks.
• Candour toward the tribunal (Rules 3.1, 3.3 and 8.4(c)). Lawyers must check AI output, including citations, before filing, and correct any false statement.
• Supervision (Rules 5.1 and 5.3). Managerial lawyers must set clear policies and ensure lawyers and staff are trained.
• Fees (Rule 1.5). Lawyers billing hourly may charge only for time actually spent, and should not bill clients for learning a tool they will use generally.

England and Wales: solicitors, barristers and judges

Barristers. The Bar Council first published its considerations on ChatGPT and similar tools in January 2024 and republished an updated version on 25 November 2025. It describes the document as a summary of principles and warnings rather than formal guidance. It accepts that barristers may use these tools within their existing duties, and it flags hallucination, bias, confidentiality, anthropomorphism and cybersecurity risks.

Solicitors. The Law Society maintains “Generative AI: the essentials”, a living guide for solicitors that it updated in autumn 2025. It covers risk assessment, data protection, client care and professional obligations, and its update references the Ayinde and Al-Haroun decisions.

The courts. In June 2025 the Divisional Court, presided over by Dame Victoria Sharp, President of the King’s Bench Division, gave judgment in Ayinde v Haringey and Al-Haroun v Qatar National Bank. Both involved fictitious authorities. The court said freely available generative AI tools are not capable of conducting reliable legal research, referred the lawyers to their regulators, and warned that more serious misuse could lead to contempt proceedings or referral to the police. For judges themselves, the judiciary publishes AI guidance that permits limited use of AI with strict safeguards on confidentiality and verification.

Elsewhere: the same principles

Regulators and courts across common law and civil law systems have reached similar conclusions. Some courts have issued practice notes or standing orders that restrict AI use in particular documents, such as affidavits or expert reports, or require lawyers to certify that AI-assisted work has been verified. In Pakistan, a Supreme Court judgment authored by Justice Syed Mansoor Ali Shah in April 2025 described AI as a facilitative tool that cannot replace judicial reasoning, and asked the National Judicial (Policy Making) Committee and the Law and Justice Commission to develop guidelines for its use in courts.

The details differ, so the practical rule is simple: before filing, check the court’s own rules. Disclosure and certification requirements now vary from court to court and sometimes from judge to judge.

The six duties that apply everywhere

Distil the guidance above and six duties remain. They are a sound working framework whatever your jurisdiction.

1. Competence

Understand what the tool does well, where it fails, and how it handles data. You do not need to be an engineer. You do need to know that general chatbots can invent cases, that outputs vary between runs, and that a tool’s confidence says nothing about its accuracy.

2. Confidentiality

Know where client information goes. Use approved enterprise accounts with training on your data disabled, check retention and sub-processor terms, and anonymise where you can. The free AI Vendor Terms Red-Flag Scanner highlights training, retention and confidentiality terms in a provider’s contract before you sign up.

3. Verification

Every citation, quotation, fact and figure must be checked against a primary source before it reaches a client or court. This single duty would have prevented every sanctions case reported to date. The sanctions cases every lawyer should know show what happens when it is skipped.

4. Supervision

Partners and managers are responsible for how juniors, trainees, paralegals and support staff use AI. That requires a written policy, approved tools, training and a review process. Courts have sanctioned every lawyer whose name was on a defective filing, not only the person who ran the prompt.

5. Candour

Never misstate the law or the facts, and if AI-assisted work contains an error, correct it promptly. Where a court requires disclosure of AI use, comply fully. Judgments consistently treat concealment far more harshly than the original mistake.

6. Fair billing and client communication

Bill for the value and time actually delivered, explain your use of AI where it matters to the client, and set expectations in your engagement terms. Clients increasingly ask about AI policies; a clear answer builds trust.

A practical compliance checklist

• Name the AI tools approved for client work, and the accounts and settings to use.
• Prohibit confidential information in unapproved tools.
• Require verification of all authorities and quotations before filing or sending.
• Keep a short record of AI use on significant matters.
• Train everyone, including non-lawyers, with real examples.
• Address AI in engagement letters and fee arrangements.
• Check each court’s disclosure or certification rules before filing.
• Review the policy at least annually, because the guidance is still moving.

For the structure of a full policy, see what an AI use policy for a law firm must cover.

The bottom line

The ethical question is not whether to use AI. Clients expect efficiency and many courts now assume some AI involvement. The question is whether you use it as a professional: understanding the tool, protecting the client’s information, verifying everything, supervising your team, and telling the truth to the court. Those duties are old. Applying them carefully to new tools is what separates the lawyers who benefit from AI from those who appear in the next judgment.