In August 2026, a federal court excluded an expert witness entirely. Not over what they said, but because their analysis relied on AI-generated citations, that turned out to be fabricated. Weeks earlier, a Florida judge found fabricated legal authority scattered across eight separate filings in a single case.
These aren’t cautionary tales from future versions of the courtroom. They’re happening right now, and at every level of the process: from self-represented litigants using generative tools to draft filings, to attorneys leaning on AI research assistants, to judges themselves exploring how AI might help manage growing caseloads.
If you’ve been telling yourself AI in the courtroom is still a “someday” problem, the last few months say otherwise. AI has likely already shown up in your court’s filings. It’s time to have a plan so you can be prepared to address it.
Courts Aren’t Waiting for Guidance Anymore. They’re Writing it Themselves
With no single federal framework in place yet, states have moved first. In the last few months alone, Connecticut, Florida, New York, and Ohio have all advanced court rules addressing AI use, including verify-and-certify requirements to formal AI-use disclosure policies. Similar movement is happening internationally, with India, the EU, and the UK introducing their own frameworks.
One federal development is worth watching closely, even though it isn’t final: Federal Rule of Evidence 707, a proposed rule that would require AI- or machine-generated evidence offered without a human expert to meet the same reliability standard courts already apply to expert testimony. It’s still moving through the federal rulemaking process (the public comment period closed in February 2026) and it remains a genuinely contested, unfinished proposal, but it signals where federal thinking is headed, even as most of the real movement right now stays state by state.
This patchwork means the specific rule your court needs to know may depend entirely on where you sit. It’s worth asking if your state has or is working on an AI-use policy for filings. If not, is one on the way?
A Quieter but Related Question: Governance?
Separate from courtroom procedure, AI tools that touch case data raise their own compliance questions, and it’s worth being precise about which rules apply. The FBI’s CJIS Security Policy governs criminal justice information specifically: it requires any vendor with access to sign a CJIS Security Addendum and put staff through fingerprint-based background checks. But most of what a court holds (family, juvenile, and mental health records, sealed and expunged files, and personal information in civil filings) falls under state statutes and court rules instead, not CJIS. California’s Rule of Court 10.430 is a good example of where that’s headed. It now requires every court in the state to have a generative AI use policy on file, including a flat ban on entering confidential or nonpublic information into a public generative AI system.
There’s also a sharper reason to ask these questions than compliance checklists alone. Earlier this year, a federal court ruled that a defendant’s conversations with a public AI tool could be used against him in court. Since the AI tool’s privacy policy allowed those conversations to be stored and shared, they didn’t get the same confidentiality protection a conversation with a lawyer would.
So before adopting any AI tool, a court should be able to get a straight answer to a few questions:
- Is our data ever used to train or fine-tune a model (theirs or a third party’s)?
- Which subcontractors or underlying model providers touch our data, and what are their retention terms?
- How long is data kept, and can it be deleted, including once the contract ends?
- Which vendor staff can access it, and are they CJIS-screened where that applies?
When AI Gets it Wrong, Someone is Being Held Accountable
The accountability conversation is moving just as fast as the rulemaking, and the sanctions landing are showing up in everyday litigation. In Los Angeles, an attorney representing State Farm was sanctioned $999.99 after opposing counsel found AI-fabricated cases, quotes, and holdings scattered across the insurer’s filings in a home-fire dispute. She told the court she hadn’t known AI had been used in preparing them, but accepted responsibility anyway as lead counsel. Additionally, a Connecticut court sanctioned a litigant for concealing an AI prompt injection attempt in a filing.
The pattern emerging: self-disclosure and transparency are being treated very differently than concealment. Accuracy, human review, and a clear audit trail are becoming the baseline expectations for both vendors and court staff.
The Vendor Landscape is Moving Fast, Which Makes it a Good Time to Ask Questions
Behind the funding rounds and acquisitions reshaping the legal AI market, there’s a more grounded story: courts are running out of people. Experienced clerks and staff are retiring, backfilling those roles is getting harder, and new state requirements, like Ohio’s e-filing mandate, are adding to the workload without adding headcount. That gap is exactly what’s drawing so much investment into court and legal AI right now: vendors see a real, unmet need, not a hype cycle. The result is a wave of funding, consolidation, and established players launching their own models or shifting from single features toward broader “AI coworker” platforms.
Our advice, the same advice we’d give regardless of who you’re evaluating, is to ask pointed questions rather than chase the newest headline: What specific problem does this feature solve for my court? Would my staff or the public actually use it? Is it something we’d pay for, or is it a box being checked because a competitor has it? Make sure it’s a feature/product your court would actually benefit from and not just the latest AI headline.
Some Courts Are Looking Outside Their Case Management Vendor Entirely
Some courts are also piloting AI tools completely independent of their CMS provider. They’re building relationships with outside AI specialists or even building in-house.
It raises a genuinely open question with no consensus yet: if your court wanted to add a public-facing AI capability today, who would be your first call? Your CMS vendor? A specialized outside partner? Your own IT team? There’s no wrong answer here, but the answer says a lot about how your court thinks about its broader technology ecosystem, and it’s exactly the kind of conversation worth having deliberately rather than by default.
Public-facing AI is Already Answering the Phone, Sort of
While the filing-and-evidence conversation gets most of the headlines, a quieter shift is happening in how courts handle the public. Multiple jurisdictions now have live, citizen-facing AI assistants handling website inquiries and answering questions like how to file a form, where the courthouse is, or how to check a case status. The Ohio House passed legislation that requires Clerks of the Court of Common Pleas and Municipal Courts to permit electronic filing.
The need is real. One clerk’s office we’ve talked with said they field literally a thousand phone calls a week from people trying to navigate the system, which is far more than their staff can keep up with. That’s the kind of volume that makes a chatbot less of a novelty and more of a practical necessity. (For anyone thinking this through, NCSC’s Court Chatbots guide is a genuinely useful, vendor-neutral starting point.)
It’s worth noting that there’s a meaningful difference between a scripted, rule-based chatbot and a fully AI-driven conversational assistant. The oversight, risk, and expectations aren’t the same for each, and it’s a distinction worth understanding before committing to either.
None of This is Cause for Alarm, but it is Cause for a Plan
Nothing above means AI is something to fear in your courtroom. It means AI is something worth being deliberate about in the same way courts have always had to be deliberate about new technology, new evidence types, and new ways people access justice.
The pace of change is real, but so is the amount of thoughtful work already happening among regulators, courts, and the vendors trying to keep up responsibly. We’d rather be honest about where things stand than pretend anyone, including us, has it all figured out.
If you’re starting to explore AI in the justice space or are further along and looking to streamline your current processes, we’ve put together A Practical Guide for Understanding and Using AI in Your Court to help you move forward with confidence.
