AI capability advances won't automatically reduce legal service costs without regulatory and institutional reforms

The legal profession anticipates significant disruption from AI systems, citing breakthroughs like GPT-4's bar exam performance as evidence of transformative potential. Legal scholars and industry leaders predict the technology will reduce costs and expand access to services that millions currently cannot afford. Narayanan, Kapoor, and Curl argue this optimism rests on flawed assumptions about how AI diffuses through institutional environments.

The researchers identify three distinct barriers that will persist even as AI capabilities improve. Unauthorized practice of law statutes and entity-based ownership restrictions prevent both consumers and lawyers from fully leveraging AI tools, creating legal liability for providers. These regulatory frameworks were designed for a different era and now act as a ceiling on AI deployment.

The adversarial nature of American litigation creates a second constraint. When both opposing counsel gain access to the same productivity tools, the competitive equilibrium shifts rather than stabilizes. Litigators can weaponize AI to increase work volume for opponents, mirroring how digital discovery expanded after digitization without reducing overall litigation costs. Transactional work exhibits similar dynamics where lawyers compete through information control.

A third bottleneck emerges from human cognition itself. Judges, lawyers, and clients require time to understand and act on legal work—a constraint that does not compress as AI accelerates task completion. If AI enables a flood of filings, courts respond by either delaying resolution or reducing review quality. This creates a ceiling on how much AI can genuinely accelerate legal processes while maintaining human oversight.

The authors acknowledge that some legal services (estate planning) operate outside adversarial frameworks and may benefit from cost reduction. However, they argue that without deliberate institutional reform addressing regulation, adversarial dynamics, and human decision-maker capacity, AI diffusion will produce abundant legal work rather than affordable legal outcomes.