The Small Firm Brief

The AI briefing for lawyers who didn't go to law school to become AI people.  ·  Issue #2  ·  May 22, 2026

This week.

The largest AI hallucination sanction in American legal history just dropped. The AI-native firm count hit 40 — they now have their own directory, in case the threat still felt theoretical. And Anthropic launched Claude for Legal ten days ago. Here's what it actually means for a small firm.

1. A federal court just issued the largest AI hallucination sanction in American legal history.

Negligence is too polite.

Two lawyers in an Oregon federal case are now more than $110,000 lighter.

Their summary judgment briefing contained 15 non-existent cases and eight fabricated quotations attributed to real authorities.

Not one or two bad cites.

Fifteen fake cases.

Eight fake quotes.

The court did not treat it as a learning-curve problem. It struck the briefs, sanctioned the lawyers, referred the matter to the Oregon State Bar, and dismissed the plaintiffs’ claims with prejudice.

There are now more than 1,400 legal decisions globally in Damien Charlotin’s AI Hallucination Cases Database. The 2026 pace is running far ahead of 2025.

Courts are done grading AI mistakes on a curve.

Pull every cited case from a real legal database before anything goes out the door. Confirm the case exists. Confirm the quote appears. Confirm the holding says what the brief says it says. Confirm it is still good law.

Research can be delegated.

Rule 11 responsibility cannot.

Not to AI.

Not to a junior associate.

Not to “I thought the tool checked that.”

$110,204.38 is a bad way to discover your citation policy was vibes in a trench coat.

2. The AI-native firm directory just hit 40. They have an index now.

When your competitors are organized enough to have their own searchable directory, the threat has moved from theoretical to structural.

Forty AI-native law firms — built from the ground up around AI workflows, flat-fee pricing, and the working assumption that the traditional law firm model is exactly the inefficiency they're selling against. Private equity is funding the buildout. The index launched in March with 27 firms. It hit 40 by April. It won't stop there.

These firms aren't coming for your judgment, your courthouse relationships, or the client who's been calling you for fifteen years. They're coming for the work that's document-heavy, price-sensitive, and repeatable.

Business formation. Standard contract review. Routine employment matters. Immigration.

If you're still billing that work by the hour, the math changes before 2027. The only question is whether you change it first or they do it for you.

3. Claude for Legal launched ten days ago. Here's an honest look at what it means for a small firm.

Anthropic launched Claude for Legal on May 12. I've spent the last ten days going through what it actually does.

The headline: 12 practice-area plugins, 20+ connectors into the software law firms run on — iManage, NetDocuments, Clio, DocuSign, Microsoft 365. The underlying model scored 90.9% on Harvey's BigLaw Bench, the industry's standard AI benchmark. Freshfields, Quinn Emanuel, and Holland & Knight announced the same day they're already using it on live matters.

For a small firm, here's what's real.

The Microsoft 365 integration is the most immediately useful piece. Claude now works across Word, Outlook, Excel, and PowerPoint as a single context-carrying agent. It reads your emails. It drafts from your documents. It carries context across a matter without you re-explaining the facts every time you open a new tool. If your firm already runs on M365 — and most do — this drops in on top of what you already pay for.

The practice-area plugins cover litigation, contract review, employment, M&A due diligence, and eight more. These are purpose-built for legal work, not repurposed from a generic AI assistant. That matters.

Here's the part nobody's saying clearly: none of this is plug-and-play.

The connectors need to be configured for your firm's environment. The plugins need context — your intake patterns, your preferred clause language, your billing structure. Out of the box, Claude for Legal is a powerful foundation. Without someone who knows how to set it up inside an actual law firm's workflow, it's expensive infrastructure that goes mostly unused.

This is the consistent pattern across every serious AI deployment in legal right now. The technology works. The implementation is what fails. And the implementation requires someone who understands both the software and how a practice actually operates day to day.

For small firms, that's not a reason to wait. It's a reason to get the setup right the first time, rather than six months from now after you've figured out what you did wrong.

Claude for Legal is the most significant AI tool aimed directly at law firms since Microsoft 365 Copilot. It deserves your attention. It will require help to deploy it well.

Claude for Legal — one AI agent across the software you already use.

If your firm runs Microsoft 365, Claude now works natively across Word, Outlook, Excel, and Teams. One agent, full matter context, no switching between tools.

Start narrow. Pick one workflow — first-draft demand letters, contract clause review, client intake summaries — and get that working before touching the more complex plugins. Prove the time savings on one matter type before you expand.

Cost: Claude Team starts at $25/user/month. The Cowork desktop app unlocks the integrations. Microsoft 365 access requires the M365 connector, which your IT setup or a consultant will need to configure.

The caveat from Issue #1 still stands: no AI tool handles citations. That step stays yours.

Prompt of the week.

Client intake summary — before a consultation.

You talk to a lot of prospective clients. Most of them describe the same three types of disputes in completely different ways. This prompt turns a client's written description into a structured intake summary you can review in 60 seconds before the call.

Paste this into Claude or Copilot. Drop the client's message at the bottom.

One more thing.

Claude for Legal is real. The integrations are real. The benchmark score is real. Big firms are already running it on live matters.

The attorneys who get the most from it over the next 12 months won't be the ones who signed up first. They'll be the ones who took the time — or got the help — to configure it for how their practice actually works.

The technology keeps getting better. The implementation problem doesn't solve itself.

Greg

Actionable AI | Less hype. More results.

P.S. Don't skip the supervision step.

Greg Yates  ·  Actionable AI for Lawyers  ·  actionableai.net  ·  [email protected]  ·  (212) 765-0685