AI breaks down Supreme Court tariff opinion into syllogisms, fallacies, and unstated assumptions in two minutes

AI Can Parse Legal Logic Faster Than Human Review

Claude and similar AI programs can break down a Supreme Court opinion's reasoning in minutes, exposing assumptions and logical gaps that might take hours to identify manually. An attorney tested this capability on a 170-page tariff case decided in February 2026, uploading the majority opinion, concurrences, and dissents to Claude Sonnet 4.6 and asking it to identify logical structure, fallacies, and unsupported assumptions.

Claude produced an 18-page analysis in two minutes.

How the Analysis Worked

The case, Learning Resources, Inc. v. Trump, involved whether the International Emergency Economic Powers Act authorized the president to impose tariffs. The majority held it did not. Rather than summarizing the opinion, Claude converted each line of reasoning into formal syllogisms-logical statements with premises and conclusions.

For the majority opinion, Claude identified that six justices reached the same result through different legal paths. The first layer relied on the constitutional principle that tariff power belongs to Congress. The second layer, used by three justices, invoked the Major Questions Doctrine. The third layer applied textual analysis, arguing that "regulate" does not mean "tax."

This structure revealed that the Major Questions Doctrine reasoning was dictum-not necessary to the decision-because the majority required additional constitutional and textual analysis to reach its conclusion.

What AI Spotlighted

Claude identified specific logical fallacies in both majority and dissenting opinions. In the majority opinion, it flagged a weak generalization: Congress used explicit terms like "duty" and "surcharge" in ordinary tariff statutes but not in the emergency statute. The majority argued this absence proved Congress did not intend to delegate tariff power. But Congress could reasonably use different language in an emergency context.

In Justice Thomas's dissent, Claude caught an anachronistic argument. Thomas cited colonial-era British royal prerogative over tariffs to support presidential authority. That reasoning ignored the Revolutionary War, which directly challenged the king's right to impose tariffs on Americans.

Claude also catalogued explicit assumptions, unstated assumptions, and unsupported leaps. The majority assumed taxing and regulatory powers are categorically distinct. It assumed a 1979 dictionary definition of "regulate" reflected what Congress meant in 1977. It inferred that 47 years of non-use meant Congress did not intend IEEPA to authorize tariffs-an inferential jump without supporting evidence.

Speed and Accuracy Trade-Offs

The speed is genuine. A lawyer reviewing 170 pages manually might reach similar conclusions, but would need hours rather than minutes. The AI output requires verification before use in a brief or filing.

One limitation: Claude worked with a single, self-contained document. It did not search for additional precedent or legislative history. The opinions themselves already contained some of the critical analysis Claude was asked to perform, meaning Claude partly synthesized what the justices had already written.

The tool's real contribution lies in making logical structure visible. Complex legal arguments often rest on a few core assumptions. Claude can surface those assumptions quickly, letting advocates and judges identify the actual points of disagreement.

A Cautionary Note

The same speed that reveals logical flaws can also generate supporting arguments for almost any conclusion. An advocate motivated to reach a particular result could use AI to produce plausible-sounding logic for positions that may not withstand scrutiny. The ease of generating arguments-even weak ones-might encourage advocates to multiply arguments rather than strengthen them.

But that risk exists with or without AI. What changes is the speed at which both weak and strong arguments can be constructed and tested.

For legal professionals, the practical value is straightforward: AI for Legal analysis can compress the time needed to understand an opinion's logical architecture. Whether that analysis is reliable depends on the same factors that govern any legal research-careful review, verification against source material, and sound judgment about which arguments matter.