Federal magistrate judge Maritza Braswell, working in Colorado's federal district court, reads legal filings from people who have no lawyer. Many can't afford one. Others have claims too small to interest an attorney. Before AI writing tools, the practical barrier to filing a lawsuit was real - a coherent brief required either law school or months of painful self-study. That barrier is largely gone, and courts are still working out what to do about it.
The MIT Technology Review covers the resulting pressure in detail, tracking how federal judges are adapting to the surge. The core tension is genuine: AI writing tools have made it significantly easier for low-income people to engage a legal system that has always been expensive and opaque. That's a real benefit. The problem is that the same tools produce filings that look professionally written but cite cases that don't exist.
When AI Invents Case Law
AI hallucinations - when a model confidently generates false information - are manageable in most contexts. In legal filings they're specifically damaging. If an AI tool writes a product review that gets a detail wrong, a reader can fact-check it in 30 seconds. If it invents a case name, a docket number, and a ruling summary, a judge has to go looking for something that doesn't exist. Federal courts already run on chronically thin resources, and phantom citations add real work with zero legal value.
Judges can't safely skip these filings. Due process requires taking each one seriously. A person filing without a lawyer (called a pro se litigant) still has constitutional rights, and a sloppy AI-drafted complaint might contain a legitimate claim buried under bad citations and invented precedents.
Disclosure Rules Are a Partial Answer
Several federal district courts have responded with standing orders requiring AI disclosure - rules that say attorneys and pro se litigants must flag when AI tools wrote or substantially edited their documents. The requirements vary: some courts demand disclosure only, others require certification that the filer checked every citation for accuracy.
These rules help at the margins. They don't solve the volume problem. The courts seeing the biggest increases are bankruptcy courts and small claims divisions, where filing fees are low enough that a long-shot complaint feels worth attempting. A filing that would have cost $2,000 in attorney time to draft now costs an afternoon with ChatGPT.
The harder problem is separating legitimate cases from noise. The person who was genuinely wronged - but whose AI-drafted complaint has three hallucinated citations mixed into an otherwise real grievance - still has rights. Judges are working through this case by case. That's not a scalable approach, and nobody in the federal court system has proposed one that is.