AI in Access to Justice and Legal Aid

In the United States, research by the Legal Services Corporation has repeatedly found that low-income households get no legal help, or not enough, for the large majority of their civil legal problems. There is no general right to a lawyer in civil cases. People therefore often face eviction, debt collection, custody and benefits disputes alone, frequently against landlords or creditors who have lawyers. The oldest technology in this space is not generative AI. It is document assembly and guided interviews. Tools such as A2J Author, Docassemble and LawHelp Interactive ask a series of plain-language questions and fill in court forms, much like tax software. Courts and legal aid groups add chatbots that answer common questions, text reminders, and triage systems that route people to the right service. Large language models add new abilities: explaining a notice in plain language, translating, summarizing a client's story for an intake lawyer, and drafting letters for staff to review. The main legal limit is the ban on unauthorized practice of law (UPL). Rules vary by state, but in general legal information is allowed: what a form is for, or when a deadline falls. Individualized advice, such as what you should argue in your case, is reserved for licensed lawyers. Some states are experimenting. Utah created a regulatory sandbox in 2020 that lets approved non-traditional providers offer legal services under supervision, and Arizona allows alternative business structures. Consumer protection law applies too. In 2024 the Federal Trade Commission acted against DoNotPay over claims about its 'robot lawyer' that the company had not backed up. A common misconception is that AI will simply replace lawyers for people who cannot afford one. So far, practice points to a supporting role: helping people get forms right and meet deadlines, and helping scarce legal aid staff serve more clients. Errors are costly, because a missed deadline can mean a default judgment or an eviction.