How we use AI.

This policy applies to generative AI — large language models, AI image generators, AI transcription, and similar tools that produce new content from input. It does not apply to features like spell-check, autocomplete, search, or word processing, even though some now use machine learning under the hood.

The policy covers AI use in everything we publish — articles, newsletter emails, YouTube videos, social media posts — and in our active client representations. We hold both to the same standard: a human attorney is responsible for what comes out, and the same confidentiality and verification rules apply.

AI tools help us with tasks that used to require a research assistant, a paralegal, or several extra hours:

• Research. Searching legal databases, surfacing statutes and regulations, finding court opinions, and pulling together background on emerging issues.
• Drafting and editing. Generating first drafts, suggesting structure, tightening prose, catching inconsistencies, and proposing edits — with an attorney rewriting and revising every published piece.
• Transcription. Converting consultation recordings, depositions, and other audio into searchable text for our case files.
• Document analysis. Reading through loan statements, court filings, agency correspondence, and other materials to identify key information and flag issues for attorney review.
• Client communication preparation. Helping us draft responses, summarize incoming questions, and prepare for meetings — with an attorney reviewing and approving every communication that leaves the firm before it reaches you.
• Internal organization. Tagging, indexing, and connecting our knowledge base so we can find what we’ve worked on before.

In each case, AI is a tool an attorney uses. The attorney decides what the tool gets to look at, what to do with the output, and whether anything it produces is good enough to publish, send, or rely on.

The list of things we don’t do with AI is shorter, but it’s the more important list:

• AI doesn’t decide what we publish. Every article, video, newsletter, and email is reviewed and approved by a human attorney before it goes out. AI tools never sit in the publication path on their own.
• AI doesn’t give legal advice. When you talk to us about your situation — whether in a consultation, in an active case, or in a client message — you are talking to a lawyer. Not a chatbot, not an AI assistant, not a model output. A lawyer reviews and is responsible for the advice you get.
• AI doesn’t replace expert judgment on YMYL topics. Articles about default consequences, garnishment, bankruptcy, and other high-stakes financial decisions are written and reviewed by an attorney who has litigated those issues. AI can help draft and edit. It can’t substitute for the experience.

The principal third-party AI tools in our workflow today are:

• Claude (made by Anthropic) — for research synthesis, drafting, editing, and document analysis.
• Perplexity — for research and source verification.
• Deepgram — for transcription of audio recordings.

We use these tools under their enterprise or API terms, which contractually prohibit the providers from training their models on our inputs. We do not use the publicly available consumer versions of these tools for client work, because those versions may use input data for training.

We also use internal platforms we built ourselves: a legal research and case management platform that holds our client work and grounds our research in the NCLC student loan treatise and primary sources, and a content workflow platform that handles our publishing operations. These are operated on infrastructure we control.

The tools we use change as the technology matures. We review this policy and our toolset at least every six months, and the “Last Updated” date at the top reflects the most recent review. When something material changes — a new tool added to our workflow, a meaningful change in how we use an existing one, a new bar opinion that affects our practices — we update sooner.

This is the section that matters most to anyone considering hiring us.

Client matter information lives in our own infrastructure. Our case management, client knowledge base, and matter records run on infrastructure we control, not in publicly available consumer AI accounts. Information you share with us as a client doesn’t get pasted into ChatGPT.

Transcripts of consultations and other recordings go through PII redaction before any model sees them. Identifying information is removed before transcripts enter our broader workflow. The redacted versions are what we work with.

When AI tools assist with analyzing client materials in active cases, we use them in configurations that don’t train on the input.

We don’t use information you share with us — through contact forms, intake forms, document uploads, or active representations — to train any AI models.

Our engagement agreements address AI use in your representation specifically. ABA Formal Opinion 512 (July 2024) and Missouri Informal Opinion 2024-11 make clear that boilerplate engagement-letter language is not enough when AI use raises confidentiality questions. When you retain us, our engagement agreement explains how AI is used in your matter and obtains your informed consent where the bar opinions require it.

These practices comply with the duties laid out in ABA Formal Opinion 512 — competence, confidentiality, communication, candor toward the tribunal, supervision, and fees — and with the guidance Missouri’s bar issued in Informal Opinion 2024-11.

Stanley Tate is the attorney responsible for the content of this site and for the work product of this firm. AI assistance doesn’t change that. If something we publish is wrong, that’s our error, not the tool’s. If something we file is wrong, that’s our error, not the tool’s. We don’t blame the model.

Every statute, regulation, agency document, and court opinion cited in our content is one we have read. AI tools sometimes generate confident-sounding but fabricated citations. The verification step catches them. If you ever find a citation on this site that doesn’t lead to the source it claims, treat it as a correction request and we’ll fix it.

AI is never the source of legal authority cited in a court filing without independent attorney verification. We are aware of the cautionary tales — Mata v. Avianca and the cases that followed — and we treat AI-generated research as a starting point that requires checking against the authority before any citation goes into a brief, motion, or other filing.

Team members work on our content and matters under attorney supervision. ABA Model Rules 5.1 and 5.3 require us to ensure that everyone under our supervision — attorney or non-attorney — operates within the rules of professional conduct. That includes how they use AI tools. We train, supervise, and review.

We commit to the following:

We label AI-generated images. If a hero image, illustration, or diagram was created by an AI image tool, we mark it. We don’t use AI to create photographs of real events or real people.

Our content is not authorized for AI training. We don’t authorize our published content — articles, videos, newsletters, or anything else — to be used to train third-party AI models without our written permission.

Our fees reflect AI-driven efficiency. ABA Formal Opinion 512’s fee duty cuts both ways: lawyers who use AI to work faster shouldn’t bill clients for hours they didn’t spend. When AI tools let us work faster, our fees reflect the time we spent — not the time the work would have taken without AI.

• Editorial Policy — how we decide what to publish
• How We Research — where our facts come from
• Disclaimers — what our content can’t do for you
• Corrections Policy — what happens when we get something wrong
• Privacy Policy — how we handle your information
• Terms of Service — the rules for using this site