Generative AI is transforming the legal profession, but the transformation isn’t (and won’t be) uniform. Some parts of the profession have embraced AI, while others have moved more cautiously. This dynamic is starkly evident even at the Supreme Court, where the AI arms race may prove particularly asymmetric. What are the implications for justice when the profession incorporates AI unevenly?
Three layers of possible asymmetry in AI adoption by judges and advocates, as well as perspectives on AI from several Justices and superstar advocate Neal Katyal, below the fold. (The tax connection? Katyal talks about his work in Learning Resources.)
As reported by Bloomberg, the Supreme Court has proved reluctant to embrace the generative AI revolution. In public comments, Justice Amy Coney Barrett disclaimed generative AI use “because it would be insecure” and affirmed that “you can trust that our opinions are not AI-generated.” Similarly, Justice Clarence Thomas recently indicated that the Court would “have some challenges with AI” and that some AI-related issues “should have been addressed with legislation.” Thomas was emphatic about his relationship with AI: “I don’t even know what that is.” These comments, of course, are consistent with the Court’s longstanding reputation for technophobia.
By contrast, Neal Katyal has touted the role of AI in his preparations for oral argument in Learning Resources v. Trump, the IEEPA tariff case that the Court decided in February. In a recent TED talk, Katyal revealed that he prepared using a “bespoke AI” developed from Harvey and “trained on every question every Justice has asked in oral argument for 25 years.” Katyal then juxtaposed his AI model’s predicted questions with the Justices’ actual questions in oral argument. According to Katyal, his AI model “predicted a possible escape route” to win over Chief Justice Roberts: “Harvey glimpsed that narrow door, I held the door open, [and] the Chief Justice walked through it.” Although Katyal emphasized his human performance, AI’s prominence in Katyal’s preparation process is notable, if perhaps overstated.
Setting aside the Internet dustup over Katyal’s TED talk, there’s pretty clearly a divide between the Justices and advocates over AI use, and this asymmetry operates along institutional, resource, and epistemic dimensions. First, there’s institutional asymmetry, where private-sector lawyers have greater license to experiment with novel technologies, while courts are constrained by Barrett’s concerns about security and legitimacy. Second is resource asymmetry. Large firms are investing heavily in AI, with cross-practice synergies that subsidize outlays directed at any specific project. Courts lack similar development and implementation budgets. This pattern isn’t new, but AI may exacerbate the divide.
Finally, AI may compound an epistemic asymmetry in which advocates have better information about the Justices than the reverse. Most Supreme Court advocates argue five to twenty cases before the Court. At present, Paul Clement has the most appearances with 124, with Katyal—a well-known repeat player—at 54. The life-tenured Justices, however, might hear 70 cases per term over more than two decades. There’s a deep body of publicly available training data on the Justices, while any in-chambers models of upcoming advocates will be comparatively thin. To the extent that AI more effectively digests these data, the Justices stand at a structural disadvantage to the advocates before them, and the Court’s substantively rangy docket likely amplifies this effect.
These asymmetries expose serious potential for systemic instability as the legal profession adapts to generative AI. If this instability is transitional, the result may be some bad outcomes before a new equilibrium emerges. But the nature of these asymmetries raises a more durable concern: AI may entrench advantages for repeat players and well-resourced litigants before courts develop institutional responses. The risk is more than technological disruption; the stakes are adjudicative legitimacy, transparency, and parity among litigants.
Related TaxProf Blog coverage:
- The Supreme Court Strikes Down Trump’s IEEPA Tariffs (Feb. 21, 2026)
- Law Prof Amicus Briefs in the Tariff Cases (Oct. 25, 2025)
