Scott Shapiro: what happens to legitimacy when law becomes code?

Artificial intelligence in law is usually sold on a promise of efficiency: faster research, faster filings, faster answers.

Scott Shapiro, philosopher of law at Yale Law School, says that framing hides the real story. The real question, on his account, is power.

Shapiro made the argument in a ninety-minute conversation published on 18 February 2026 in the European Corporate Governance Institute's New Law Order series.

The core of the argument#

Shapiro's starting point comes from his own theory of law: law is a social technology that coordinates human behaviour at scale.

If that is right, then tools that can interpret rules, test edge cases and produce persuasive legal analysis do not merely speed the work up. They change who can navigate the system and who is buried by it.

A dark symmetry#

The most striking part of the conversation shows that this shift runs both ways.

Tools that make it easier for people to comply with their legal obligations also make it easier for well-resourced actors to evade them. Shapiro describes the compliance engine and the exploitation engine as the same machine.

The question that follows: what happens to the balance of justice when the best "law hacker" is a machine?

Two different dangers#

Shapiro draws an important distinction between generative models and rule-based systems. Their problems differ.

Generative models can speak with an authoritative tone and still be wrong — whether by accident or design. In law, a hallucination means invented case law or a citation to a provision that does not exist.

Rule-based systems do not carry that problem; their logic is sound. But they raise another question: what happens when the logic is correct but the reasoning can no longer be followed by a human being?

Shapiro's answer is uncomfortable. At that point "legal reasoning" stops being a reason and becomes an appeal to authority. The decision may be right; but if nobody can follow why it is right, there is a serious problem of legitimacy, transparency and trust.

Why this man?#

Shapiro is Charles F. Southmayd Professor of Law at Yale Law School and a professor of philosophy. His work runs from the philosophy of law to international law, from cybersecurity to artificial intelligence.

The planning theory he developed in Legality (2011) treats law as a system of plans: a mechanism for producing plans that lets us coordinate in large groups without having to settle complex and contested moral questions first.

The Internationalists (2017), written with Oona Hathaway, defends the long-run effect of the 1928 Kellogg-Briand Pact, which outlawed war.

Fancy Bear Goes Phishing (2023) tells the history and ethics of hacking through the stories of five major cyberattacks. Shapiro's present assessment of artificial intelligence reads as a continuation of that book: in both, the question is the same — does the person who can break a system stand outside its rules, or are they the person who understands them best?

The context#

Shapiro's warning connects to a long-running argument in the philosophy of law: rules or judgement?

The rule of law defends the rule against arbitrariness. But a rule cannot foresee every case; applying it requires judgement. Artificial intelligence presses on that balance from both sides: it can automate the application of the rule, and it can make the grounds of the judgement invisible.