The brief filed in a federal district court in Manhattan in 2023 contained citations to six cases that did not exist. The lawyer who filed it, using an AI legal research tool that had fabricated plausible-sounding but entirely fictional precedents, told the judge he had not realized that artificial intelligence could generate false information. The case became a cautionary tale in bar association newsletters and law school seminars across the country. It was also, in retrospect, an early data point in a transformation of legal practice that was already well underway and has only accelerated since. The ways artificial intelligence is being used in American law now fall into several broad categories. Document review, the laborious process of reading through thousands or hundreds of thousands of pages of discovery materials to identify relevant information, has been transformed by machine learning tools that can process far more material in far less time than human reviewers. E-discovery technology has been developing for over a decade, but the latest generation of tools represents a substantial capability leap. Contract analysis is another area of rapid change. Large language models can now review commercial contracts, flag unusual provisions, compare terms against standard market practice, and summarize key obligations at speeds that human lawyers cannot approach. Law firms with large transactional practices have integrated these tools into their workflows, and the time billed for certain categories of document review has fallen correspondingly. Legal research, the task at the center of the 2023 hallucination scandal, is where the risks are most visible and where lawyers have been most cautious. The established legal research platforms, Westlaw and Lexis, have developed AI tools that operate within their verified databases and are therefore less susceptible to fabrication. But the broader availability of general-purpose AI tools means that some lawyers and many self-represented litigants are using systems that do not have the same guardrails. The implications for access to justice are genuinely contested among legal experts. On one hand, AI-powered legal tools have the potential to make basic legal assistance available to people who cannot afford lawyers, helping them understand their rights, draft simple documents, and navigate bureaucratic processes that have historically required professional help. Several nonprofits and legal aid organizations are developing AI tools specifically for this purpose. On the other hand, the availability of AI tools that produce plausible-sounding but potentially inaccurate legal analysis creates new risks for unrepresented individuals who may not have the expertise to evaluate what the tools are telling them. A landlord-tenant dispute in which the tenant relies on AI-generated advice that misstates the applicable law has concrete human consequences. In the courtroom itself, AI is beginning to appear in ways that raise more fundamental questions about procedural fairness. Algorithmic risk assessment tools have been used in criminal sentencing and bail decisions for years; the debate about whether their outputs introduce systematic bias continues. Facial recognition technology has been used in criminal investigations in ways that have produced wrongful identifications. And evidence generated or processed by AI systems is beginning to create questions about authentication and reliability that existing evidentiary rules were not designed to address. The law's characteristic conservatism, its tendency to move slowly and to favor established procedures, has meant that the legal system is absorbing the AI revolution more gradually than some other fields. But the pace is faster than it has ever been for any previous technological disruption. How courts, bar associations, and legislatures respond over the next few years will shape how these tools are used in one of the most consequential domains of public life.