Artificial intelligence is no longer a speculative topic for United States attorneys. Firms of every size, from national practices to solo shops, are already using AI tools for concrete, practical tasks. Adoption is neither universal nor uniform, but it is real, and it is accelerating. The useful question is no longer whether AI belongs in a law practice, but where it genuinely helps, where it does not, and what professional duties govern its use.
This article takes both halves of that question seriously. The first part looks at where AI is delivering real value for US attorneys today. The second part turns to the professional best practices a careful attorney keeps in mind when putting these tools to work: protecting client confidentiality, understanding how a tool handles data, keeping a human in the loop, and verifying output. The goal is an accurate, grounded assessment rather than either breathless promotion or reflexive fear.
Where AI Genuinely Helps US Attorneys
AI applications in law fall into several categories, each at a different level of maturity. Understanding which is which helps an attorney decide where to start and where to remain cautious.
Drafting
Drafting is one of the clearest early wins. AI tools can produce a first draft of routine correspondence, a client update, a demand letter, or an internal memo far faster than starting from a blank page. Used well, this does not replace the attorney's judgment. It removes the friction of the initial draft so the attorney can spend time on the parts that require legal skill: the analysis, the strategy, and the tailoring to the specific client and matter.
The important discipline is that a draft is a draft. Every AI-produced document must be read, corrected, and made the attorney's own before it goes anywhere near a client or a court. The tool accelerates production. It does not transfer responsibility.
Legal Research Assistance
Research was one of the first areas where AI showed practical value. Modern legal research assistants can take a question in plain language, surface potentially relevant authorities, summarize holdings, and suggest lines of argument. For an attorney facing an unfamiliar issue, this can meaningfully shorten the path from question to a working understanding.
The value comes with a firm caveat. AI research tools are not infallible. They can miss controlling authority, misstate a holding, or, in the case of general-purpose tools, generate citations to cases that do not exist. The attorney must independently verify every authority and every proposition before relying on it. AI shortens the research process; it does not replace the attorney's obligation to confirm that the law says what the tool claims it says.
Documentation and Memos to File
Documentation is an area where AI is quietly valuable and often overlooked. Turning a spoken account of a client call or meeting into a clean, structured memo to file is exactly the kind of well-scoped, high-volume task where AI shines. The attorney narrates what happened, and the tool produces an organized memo with a summary, key points, client instructions, and action items.
This matters for more than convenience. Contemporaneous, well-structured documentation keeps you organized, helps you keep clients properly informed, and, when a relationship sours, is frequently an attorney's best defense to a malpractice or fee dispute. A memo written minutes after a conversation carries far more weight than a reconstruction assembled months later. Anything that makes it faster to capture that record while the details are fresh has real professional value.
Document Review
In discovery and due diligence, AI-assisted review has matured considerably. Tools can classify, cluster, and prioritize large document sets, flag likely-relevant or privileged material, and surface patterns a manual review might take weeks to find. For document-heavy litigation and transactions, this is among the most established and defensible uses of the technology.
Even here, supervision remains the attorney's job. The tool narrows and prioritizes; the lawyer remains responsible for the decisions that follow from the review, including privilege calls and production judgments.
The Professional Picture: Good Practice With AI
Enthusiasm for AI has to sit alongside the professional responsibilities every attorney already carries. Those responsibilities are not new, but they apply to AI in specific and important ways. The themes below, competence, confidentiality, verification, and supervision, are simply good practice applied to a new kind of tool. Thinking around AI is still developing, so it is worth staying current with how your own professional community is approaching these tools.
Understanding the Tools You Use
Competent work has always included understanding the tools you rely on, and as a matter of good practice that extends to keeping a working sense of the benefits and risks of the technology in your practice. For AI, the practical meaning is straightforward: an attorney who uses an AI tool should understand, at a working level, what the tool can and cannot do, where it tends to fail, and how to check its output. Using a tool you do not understand well enough to supervise is itself a problem, not a shortcut.
Confidentiality and Vendor Data Handling
Protecting client confidentiality is one of the oldest commitments in the profession, and a careful attorney makes reasonable efforts to prevent client information from being disclosed without authorization. When client information is entered into an AI tool, that concern extends to the vendor. Before adopting a tool, an attorney should understand where the data goes, who can access it, how it is stored and secured, and, critically, whether the provider uses submitted content to train its models.
Concrete questions worth asking any vendor include whether data is encrypted in transit and at rest, whether inputs are used for model training, whether the provider offers contractual data protection terms, and how the tool measures up against the privacy laws that apply to your clients' information. A tool that encrypts data with strong encryption such as AES-256 and contractually commits not to use client content to train its models sits far more comfortably with a confidentiality-minded practice than a consumer tool with vague or permissive terms. Confidentiality is not a checkbox; it is an ongoing exercise of care.
The Duty to Verify Output
Nothing matters more in the AI context than verifying what the tool produces. AI systems can produce fluent, confident, and entirely fabricated content, a phenomenon commonly called hallucination. There have been widely reported instances where attorneys filed briefs containing citations to cases that did not exist, generated by an AI tool and submitted without being checked. The fallout has included public embarrassment, lost credibility with the court, and real harm to client interests and professional reputations.
The lesson is not that AI is unusable. It is that AI output is a starting point, never a finished product. Every citation must be pulled and read. Every factual assertion must be confirmed against a reliable source. Every quotation must be checked against the original. The attorney who signs a filing vouches for its contents, and no AI tool absorbs that responsibility. Verification is the price of using the technology, and it is not optional.
Treat every AI output the way you would treat a draft from a brand-new associate you have never worked with: potentially useful, occasionally brilliant, and never something you would file, send, or rely on without reading and checking it yourself.
Keeping a Human in the Loop
A useful way to think about an AI tool is the way you would think about any assistant working under your name: an attorney who uses one should supervise its work with the same care applied to a nonlawyer assistant. That means putting reasonable measures in place to keep the tool's use consistent with your professional responsibilities, including confidentiality and competence, and reviewing its output before you rely on it. The responsibility for the work product stays with the attorney, regardless of how much of the initial effort a tool performed.
The bodies that guide the profession have been converging on the same message, pulling together these threads of competence, confidentiality, communication with clients, reasonable fees, and supervision. The message is not that AI is off limits. It is that the responsibilities a careful attorney already honors continue to apply. That is the right mental model: AI does not create a new rulebook, it applies familiar good practice to a new tool. It is worth staying current, since thinking in this area is developing quickly.
What Attorneys Should Consider Before Adopting a Tool
For a firm weighing whether and how to bring AI into its practice, a small number of questions separates responsible adoption from reckless adoption:
- Scope. What specific task is the tool for? Narrow, well-defined tasks such as drafting a memo to file or prioritizing a document set are far safer starting points than open-ended reliance.
- Confidentiality. Where does client data go, how is it secured, and is it used to train the vendor's models? Get the answers in writing.
- Verification workflow. Who checks the output, and how? A tool is only as safe as the review process wrapped around it.
- Competence. Does the team understand the tool well enough to supervise it? If not, training comes before deployment.
- Client communication. Does the representation or the circumstances call for telling the client that AI is being used? When in doubt, err toward transparency.
- Keeping current. Thinking on AI in professional settings is evolving quickly, so it is worth following how your own professional community is approaching these tools.
A sensible adoption path is to start with a low-risk, high-frequency task where the payoff is clear and the downside is contained, build a verification habit around it, and expand from there as the team's understanding grows. Documentation and internal drafting are natural first steps precisely because they are bounded, the attorney reviews the output before it leaves the firm, and the efficiency gains are immediate.
The Realistic Outlook
AI is not going to replace attorneys, and it is not a passing fad. The realistic outlook is that AI becomes a standard part of the toolkit for well-scoped tasks, much as legal research databases and document management systems did before it. The attorneys who benefit most will be those who treat it as a capable but fallible assistant: quick to draft, tireless at review, and always in need of supervision.
The duties that govern its use, competence, confidentiality, verification, and supervision, are not obstacles to adoption. They are the conditions that make adoption responsible. An attorney who understands the tool, protects client information, checks the output, and supervises the work can capture the genuine efficiency AI offers without compromising the professional obligations that define the practice of law.
Conclusion
AI is changing US legal practice in real ways: faster drafting, more efficient research and document review, and dramatically less friction in day-to-day documentation. Those gains are genuine and worth capturing. They come with responsibilities that are equally genuine and rooted in the good practice every attorney already follows: understanding the tools you use, protecting client confidentiality, verifying AI output yourself rather than delegating that judgment, and supervising the work a tool performs.
Used within that framework, AI is a powerful ally rather than a liability. Tools built for law, with strong encryption, clear commitments not to train on client content, and a workflow that keeps the attorney in control of the final product, make it far easier to stay on the right side of those duties. If your practice is looking for a low-risk place to start, structured documentation is one of the most defensible entry points. See how Lex Protocol turns a spoken account into a structured memo to file while keeping the attorney firmly in the driver's seat.