AI is a tool for lawyers, not a replacement for judgement
Artificial intelligence is now woven into legal work — from contract review to research to drafting — and most regulators permit it. The consistent message from bar associations and courts is that AI is a productivity tool, and the lawyer remains fully responsible for the output. Duties of competence, confidentiality, candour to the court, and supervision do not disappear because a machine helped produce the work.
What AI does well in legal practice
Used carefully, AI can meaningfully speed up several tasks:
- First-pass contract review — flagging unusual clauses, missing provisions, and risk terms for a lawyer to examine.
- Summarising — condensing long documents, depositions, or case files into briefings.
- Drafting starting points — producing first drafts of correspondence, clauses, or memos that a lawyer then edits and verifies.
- Research assistance — helping explore an area or generate search terms, provided every authority is checked against a primary source.
- Document organisation — clustering, tagging, and extracting structured data from large document sets.
In each case the value is acceleration. AI gets you to a reviewable draft faster; it does not deliver finished, citable legal work on its own.
The hallucination problem with case law
The single most damaging failure mode is fabricated authority. General-purpose models can generate citations, case names, holdings, and quotations that look entirely real but do not exist. There are now multiple reported instances of lawyers being sanctioned for filing briefs containing fictitious cases produced by AI. The rule is simple and non-negotiable: never cite anything an AI produces without verifying it in a primary legal database. Treat AI output as an unverified draft, not a source.
Confidentiality, privilege, and data risk
Legal work is built on confidentiality, and AI introduces new data-handling risks. Pasting privileged or sensitive client information into a consumer chatbot may breach confidentiality obligations and, depending on the tool and jurisdiction, raise questions about privilege. Before any client data goes near an AI tool, lawyers should use enterprise-grade products with a data-processing agreement, explicit no-training guarantees, access controls, and clear retention terms. When in doubt, anonymise or keep the data out.
Professional conduct and disclosure
Beyond confidentiality, lawyers must consider their duty of competence — understanding the tool’s limits well enough to use it responsibly — and the duty of candour, which means not misleading a court with unverified AI output. Some courts now require disclosure of AI use or certification that AI-generated content has been checked. Billing practices also need care: charging for hours saved by automation can raise ethical questions. Always check the current rules of the relevant bar and court.
The bottom line
Lawyers can use AI, and increasingly should, to work faster — but only as a supervised assistant. Verify every citation and factual claim, protect client confidentiality with appropriate tools, follow professional-conduct and disclosure rules, and keep a qualified human firmly in charge of judgement and final responsibility. Used that way, AI is a genuine accelerator; used carelessly, it is a fast route to malpractice.