The vast availability of generative artificial intelligence has created a fundamental question at the intersection between emerging technology and privilege doctrine: does a litigant waive attorney-client privilege or work-product protection by using generative artificial intelligence to analyze legal issues or develop case strategy when such action is taken with the intent of communicating the results to an attorney or using the results during litigation? The question moved from hypothetical to actual in United States v. Heppner, the first reported decision to conclude that such computer-generated communications are unprotected under both doctrines.

This Article challenges the reasoning underlying Heppner, proposes alternative results based on existing privilege doctrines, and warns courts not to treat generative artificial intelligence as though it were a human recipient rather than what it actually is—a computer-based tool. To be clear, this Article does not argue for a blanket artificial intelligence privilege. Rather, drawing on the historical development of the attorney-client privilege and work-product doctrine, as well as a century of jurisprudence adapting confidentiality and privacy principles to technological advancements, this Article contends that a party's communications with a generative artificial intelligence tool should not automatically forfeit the confidentiality protections that underlie the attorney-client privilege and work-product doctrine.

Central to this analysis is the United States Supreme Court's June 2026 decision in Chatrie v. United States, which significantly expands modern privacy jurisprudence in the context of evolving technology and personal data held by third-parties. This Article is at the forefront of examining how Chatrie's privacy reasoning informs the application of attorney-client privilege and work-product doctrine to generative artificial intelligence, demonstrating the same principles that support reasonable expectations of privacy in location-tracking technologies held by third-parties—namely, the "revealing nature" of such data and the fact that such data is not "truly shared" with another person as that term is traditionally understood—likewise support preserving confidentiality when litigants use generative artificial intelligence to do legal analysis in preparation of speaking with an attorney or in anticipation of litigation.

It is a sunny afternoon in rural Georgia. A young Black man goes for a jog. On his route, he crosses the train tracks and heads into a predominantly White neighborhood. As his sneakers rhythmically hit the pavement, he hears the rev of a pickup truck behind him. Shortly after, he hears a second truck. The young jogger picks up his pace, but so do the trucks. The trucks chase him and eventually surround him. Two White men get out with shotguns. One stands ready, barrel raised. The other approaches the jogger and fires a bullet into his chest.

That is the story of Ahmaud Arbery’s murder in February 2020. During jury selection at his killers’ trial the following year, defense counsel removed almost every Black juror eligible to serve. The prosecutor made a Batson challenge, but the judge denied it. Despite finding there had been “intentional discrimination,” the judge said defense counsel had provided “race-neutral” reasons for removing the Black jurors. Although the judge made clear he did not “feel like” those were the true reasons, he said Batson’s “limitations” prevented him from finding otherwise.

For forty years, the Batson test has consistently failed to remedy the problem it attempted to solve—racial discrimination in jury selection. This Article explores the well-known failures of Batson and boils them down to one word—subjectivity. It discusses solutions that have been proposed to fix the problem, including eliminating peremptory challenges entirely. However, this Article explains that peremptory challenges play an important role in our adversarial system, and we do not need to throw the proverbial baby out with the bath water to fix this problem. Rather, this Article analyzes how four states have recently created objective solutions to Batson’s subjective problem, fulfilling the promise of Equal Protection that Batson aspired to achieve but fell short of delivering.

Racist rhetoric has no place in the American courtroom, regardless of which side of the aisle invokes it. Although courts and scholars have long condemned prosecutors’ use of racial stereotypes and racially charged arguments to secure convictions, considerably less attention has been paid to the use of the same rhetoric by criminal defense attorneys seeking acquittals. This Article argues that defense attorneys should be held to the same legal and ethical standards they demand of prosecutors: advocacy designed to exploit racial prejudice, stereotypes, or implicit bias is improper.

This Article examines how racially coded language, “othering,” and implicit appeals to racial stereotypes can influence juror decision-making even when an attorney never expressly references race. It explores the psychological and legal consequences of racial stereotyping in the courtroom and argues that vigorous representation of a criminal defendant neither requires nor justifies advocacy calculated to trigger racial bias.

Building on that analysis, this Article proposes concrete measures for preventing such rhetoric from influencing jury verdicts. Among other reforms, it argues that courts should require any party intending to make explicit or implicit race-related arguments to first seek a pretrial ruling through a motion in limine; employ appropriately timed limiting and anti-bias jury instructions; and intervene sua sponte when attorneys improperly appeal to racial prejudice. Ultimately, the Article proposes a simple but important principle: courts should evaluate racially prejudicial advocacy based not on who makes the argument, but on what is said and its potential to improperly influence the jury.