Law firms can use generative AI to assist with research, drafting, summarization, and document-heavy work, but lawyers remain responsible for the advice, filings, and client information involved. Safe adoption depends on understanding each tool, setting firm controls, verifying outputs, and following the rules and client terms that apply to the matter.
What AI can—and cannot—do in legal practice
Law firms already use AI-based technology in some workflows, including technology-assisted review in e-discovery. Generative AI extends the range of tasks where software can produce or reorganize text from prompts. It may help with first drafts, summaries, document organization, or research workflows, but these are potential uses—not guaranteed productivity gains or substitutes for legal analysis.
As an Amazon Associate I earn from qualifying purchases.
The practical distinction is between assistance and responsibility. A tool can generate a useful starting point; a lawyer must decide whether it is accurate, complete, appropriate to the matter, and safe to use. The American Bar Association’s Formal Opinion 512 applies existing professional duties to generative AI. As the ABA put it in its July 29, 2024 announcement, lawyers and firms using generative AI must “fully consider their applicable ethical obligations,” including competent representation, protecting client information, communicating with clients, and charging reasonable fees consistent with time spent using AI.
Recommended Free Tools
Formal Opinion 512 interprets the ABA Model Rules. Those rules are models, not automatically the governing rules in every jurisdiction. State rules, tribunal orders, engagement agreements, client instructions, and other applicable requirements may add or differ.
#1 Best Overall
Which professional duties matter most?
Competence: understand the tool and examine its work
A lawyer should have a reasonable understanding of a tool’s capabilities and limitations before relying on it. That includes checking whether it covers the relevant jurisdiction and source types, what it can and cannot retrieve, and how its output should be verified. A polished answer is not evidence that it is correct or complete.
Confidentiality: assess data handling before entering client material
Do not assume that a consumer-facing interface or an enterprise label, by itself, establishes that client information is protected in a way the firm can accept. Before entering confidential information, review the provider’s applicable terms and data practices. The questions include what is retained, who can access submissions, how long information remains available, and whether submitted data may be used for further model development. Follow the firm’s approved-tool rules and any client or matter-specific restrictions.
Firm controls should also account for who can use a tool and what information they can reach. Access design needs to respect confidentiality obligations and conflicts screens rather than treating every user or matter as interchangeable.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minutePC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Communication: decide whether disclosure is required in context
There is no blanket conclusion that every use of AI must be disclosed to every client. Check the engagement agreement, client instructions, the significance of AI to the representation, and the rules or guidance in the governing jurisdiction. If a client has imposed a specific restriction or the use is material to the representation, that context may affect what communication is appropriate or required.
Supervision: include staff and other agents in the controls
A firm’s responsibilities do not end with the lawyer who signs off on a matter. Lawyers must supervise nonlawyer staff and agents involved in client work, including those using AI tools. Policies should identify permitted uses, the human review required, and how users should raise concerns. Training should cover limitations, verification, client-data handling, and prohibited uses.
Candor and filings: verify authorities and quotations
Lawyers remain responsible for claims, citations, and quotations submitted to a court. Before filing, check each authority and quotation against authoritative legal sources, and confirm that the cited material supports the proposition for which it is offered. A tool’s citation or summary is a lead to verify, not a replacement for checking the source or the rules of the particular tribunal.
Rank #4
Fees: bill for the work actually performed
Under ABA Formal Opinion 512’s discussion of Model Rule 1.5, fees must remain reasonable. For hourly work, the bill should reflect time actually spent; a lawyer should not charge for time not spent merely because a task used to take longer. The cost of learning how to use a tool generally is not passed on to the client. The applicable rule, engagement terms, and fee arrangement still matter.
Why human review remains essential
A 2024 empirical study reported hallucinations in 17% to 33% of tested outputs from two legal research tools in its particular benchmark. That range describes the study’s products, prompts, and test setup; it is not a universal error rate, a prediction for every task, or a claim about the performance of any current product in all situations.
For legal work, verification should focus on the substance as well as the citation format. Check that a case exists, that it remains relevant, that the quotation is exact, and that the cited authority supports the legal proposition. Where a tool produces a summary, read the underlying authority when the issue matters to advice or a filing.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How widespread is law-firm adoption?
Thomson Reuters Institute reported that the share of legal organizations incorporating generative AI rose from 14% in 2024 to 26% in 2025. These are publisher-reported survey findings, not a census of law firms or lawyers. The publisher’s April 2025 release described research surveying nearly 1,800 global professionals across legal, tax, accounting, corporate risk, and government sectors; that full sample should not be described as lawyers or law firms.
The figures indicate increased reported adoption, not that every organization uses AI in the same way, has approved the same tools, or has achieved a particular result.
How should a firm evaluate an AI tool?
Evaluate the tool against the firm’s actual matters and workflows rather than relying on a broad product label or a vendor claim. The reviewed evidence does not establish a current vendor-by-vendor comparison, specific security certifications, or current product terms. A firm should independently assess the following before approving a tool:
- Coverage: Does it cover the relevant jurisdiction and authoritative sources for the intended work?
- Traceability: Can users follow citations to source material and verify it efficiently?
- Data terms and controls: What are the confidentiality, retention, access, and data-use terms? Can the firm limit access appropriately?
- Matter boundaries: Can permissions and ethical walls be preserved across clients and matters?
- Workflow fit: Does it integrate appropriately with the firm’s document, research, and case processes?
- Review and supervision: What audit trail is available, who must review outputs, and what training will users need?
- Total cost and billing impact: What does the tool cost to operate, and how will its use fit the fee arrangement without charging for time not spent?
A practical sequence for adoption
- Choose a bounded use case. Start with a limited task and define what an acceptable result looks like, including what errors would make the output unusable.
- Classify the information involved. Identify what data the task requires and what client or matter information must not be entered.
- Review the provider’s terms and controls. Check retention, access, confidentiality, and any use of submitted data before approving the workflow.
- Approve the tool centrally. Assign ownership, set access rules, define escalation routes, and make review responsibility explicit.
- Train everyone who will use it. Cover the tool’s limitations, verification expectations, client-data handling, and prohibited uses.
- Require lawyer review at the right point. A responsible lawyer should review outputs before they become client advice, are delivered to a client, or are filed with a court.
- Monitor and revisit approval. Track quality, security, cost, and relevant rule changes; reconsider approval when the provider changes its product or terms.
The ABA’s risk-management recommendations offer approaches firms can adapt; they are not binding universal standards. A US-centered ABA analysis is a starting point, not a substitute for checking current state rules, tribunal requirements, and client obligations for the particular work.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
