On Monday, a federal judge in Mississippi removed all four attorneys from a case. Both sides. Plaintiff counsel and defense counsel. The reason: both submitted briefs citing cases that don’t exist.

Senior U.S. District Judge Sharion Aycock revoked pro hac vice admissions, barred the out-of-state lawyers for two years, and imposed monetary sanctions on all four. The local counsel admitted they signed the briefs without reading them. The out-of-state counsel admitted the fake citations came from AI.

Same case. Same day. Both sides. I can’t think of a better argument for getting governance in place before you let anyone at your firm touch these tools.

Five companies want to be your firm’s AI layer

The timing is almost funny. Attorneys are getting bounced from cases for using AI carelessly, and the legal tech industry is simultaneously spending billions trying to get you to use more of it.

Anthropic launched Claude for Legal in May with 20+ connectors to platforms like iManage, NetDocuments, Relativity, and Thomson Reuters. Twelve practice-area plugins. They want Claude wired into every tool your firm already runs.

OpenAI powers Harvey (which has raised $300M+) and is the engine behind Microsoft Copilot. So they’re already in your Word, Outlook, and Teams whether you’ve thought about it that way or not.

Microsoft is bundling Copilot into new M365 SKUs starting July 1. Higher price, AI included. Their bet is simple: Copilot lives where you already spend 8 hours a day, so why would you go elsewhere?

Clio acquired Jurisage last week, a Canadian legal AI data company. They called it “a foundational investment in the future of legal AI.” Translation: Clio is building its own AI into the practice management platform thousands of small firms already use.

And a company called Eve just launched “EveOS” today, an AI system designed to run a plaintiff firm’s entire case lifecycle.

Five different companies, five different answers to “how should AI work in a law firm.” All of this in the last 30 days.

So how do you actually pick?

If you’re a 20-person firm, the “should we use AI?” conversation is over. You’re going to use it. But which ecosystem do you commit to? These platforms don’t all talk to each other. Claude’s connectors work with iManage. Copilot works with SharePoint. Clio is building its own thing. Pick wrong and switching costs pile up fast.

Pick any of them without governance and you’re one hallucinated citation away from Judge Aycock’s courtroom.

A few things worth sorting out before you evaluate platforms:

Where does your data live? If you’re a NetDocuments shop, Claude’s connectors matter. If everything is in SharePoint, Copilot has the home field advantage. If you run Clio, their native AI will have the deepest context on your matters and billing. The answer to “which AI?” usually starts with “where’s our data?”

What does “wrong” look like? In Mississippi, wrong was fake case citations. At your firm, wrong might be AI pulling privileged information from another client’s folder into a draft. Or summarizing a deposition inaccurately. Or generating a client email in the wrong tone. You need to define your failure modes before you can prevent them.

Who actually reviews the output? Those Mississippi local counsel signed briefs without reading them. That’s a workflow failure, not a technology failure. Whatever tool you adopt, someone specific needs to own the review step. Not “everyone should check,” but an actual named person with that responsibility.

What do courts require now? Florida’s new rule takes effect June 15: every filing must certify that “legal authorities identified exist and are accurately cited.” Five states have some version of this. Iowa doesn’t yet, but the ISBA’s annual meeting later this month includes CLE sessions on AI in practice. It’s coming here too.

Governance first, then platform shopping

The firms getting sanctioned didn’t pick the wrong vendor. They picked a vendor and skipped the boring governance work.

Before you evaluate platforms, you need a few things in place: an acceptable use policy defining which tasks AI can assist with; a review process that names who checks AI output and when; a system for flagging AI-generated content in your documents; and a sandbox to test things in before using them on live client work.

The platform landscape will look different in 12 months. The governance you build now is what keeps your firm off a sanctions list while the dust settles.

Artech Solutions helps Iowa law firms and professional services companies evaluate, deploy, and govern AI tools. If your firm is working through these questions, let’s talk.