Legal AI Ethics: Three Things to Keep in Mind

Lawyers are using AI. In the United States, Clio’s 2026 report says 71% of solo practitioners and 75% of small firms use it for legal work. I suspect many Indian lawyers use it too, perhaps for individual tasks as they arise, though I do not have figures for India.

Using AI raises three familiar questions: Have you told the client? Have you checked the work? And have you kept the client’s information safe?

Tell the client what you are doing

It is sensible to explain to clients how your practice uses AI and what safeguards you apply. A short explanation in the engagement letter gives clients an opportunity to ask questions before work begins. Check whether applicable rules, client instructions or the circumstances of the matter call for disclosure or consent.

A client may ask whether using AI will reduce the fee. That is a fair question. Be clear about how the work will be billed and discuss any concerns. AI may save time, but the lawyer still brings judgment, experience and responsibility to the work.

Check the work, including the cases

A polished first draft is still a first draft. Complex legal documents need several rounds of review and revision, whether the first version came from a lawyer or an AI system. Check the facts, the reasoning and the final language.

Legal research deserves special attention. AI can confidently produce a case that does not exist, a quotation that was never written, or a real judgment that says something quite different. In Pooja Ramesh Singh v Jammu and Kashmir Bank, the Supreme Court of India set aside tribunal decisions that relied on fabricated authorities.

An unreliable citation can waste court time, delay a matter and harm the client. It may also lead to costs or professional consequences. The lawyer remains responsible for checking the authorities before relying on them.

A responsible legal AI workflow should make checking easier by providing links to the original sources. Open or download them, and check that each case exists, the quotation is accurate, the proposition is supported and the law remains current.

I have developed ProAI Legal to bring research and drafting into a more organised workflow. The aim is to make review easier. The lawyer must still check the work.

Protect the client’s information

Confidentiality did not begin with AI. Lawyers have long had to protect files, computers, conversations and documents, including when using email, cloud services and other outside providers. AI calls for the same care, with attention to how a particular service processes, retains and uses the information supplied to it.

For example, Notion says that, by default, neither it nor its AI providers use customer data to train models. It also says AI providers for Free, Plus and Business plans retain data for 30 days or less. On the Enterprise plan, Notion offers zero data retention for most AI models, although some models retain data for a limited period and require the workspace owner to opt in.

Before uploading anything, consider what is really needed. Remove identifying details where practical, restrict access, use strong authentication and check the service’s terms, retention practices, subprocessors and data location. A training opt-out does not necessarily mean that information is never processed or retained.

For particularly sensitive matters, stronger controls—or not using an external AI service at all—may be appropriate. If your client is a government agency, for example, its security rules and restrictions on providers, processing locations and cross-border transfers may determine which service you can use. So, check applicable law, client instructions and contractual restrictions.

Conclusion

The use of AI in legal practice is inevitable. As the Supreme Court made clear in the judgment discussed above, the rightful use of AI is not prohibited. The important thing is to exercise care at every stage and keep the lawyer’s judgment at the centre of the work.

A responsible legal AI system should break the work into manageable steps, so the lawyer can review, verify and approve the results at each stage. That is the approach behind ProAI Legal: it organises the workflow to support human review and verification, rather than treating a first answer as the final product. The lawyer remains responsible for the decisions and the finished work.


Research


Discover more from EdTeX Technologies

Subscribe to get the latest posts sent to your email.

Leave a Reply