AI in the Courtroom: What Happens When Both Lawyers Are Using the Same Tool?

The logic Hormozi is pointing at

Source: X - Alex Hormozi

Alex Hormozi posted a short observation this weekend that is getting traction: if AI replaces judges and juries, and attorneys are already using it to construct arguments on both sides, then the whole adversarial process eventually collapses into fact submission and output. No theater. No persuasion. Just inputs and a verdict.

It is a deliberately provocative framing, but it is not an unreasonable extrapolation from what is already happening. Law firms are actively deploying AI for legal research, brief drafting, jury selection analysis, and argument preparation. The American Bar Association's 2025 Year-Two Report on AI in law noted a shift in professional attitude from whether to use AI to how - which means adoption is no longer a question of if.

How far AI has already moved inside courts

Judges are already using AI for administrative tasks - summarizing documents, drafting routine orders, building case timelines. Over 60 jurisdictions in the United States use AI-powered pretrial risk assessment tools to inform bail decisions. China unveiled an AI judicial platform in late 2024 designed to classify legal information and reduce case processing burden on human judges.

In February 2026, researchers at the University of North Carolina School of Law ran an experiment where ChatGPT, Claude, and Grok were set up as a simulated jury. The three AI platforms deliberated together, changed positions based on each other's reasoning, and after 13 minutes converged on a not guilty verdict - one that differed from the actual human verdict in the real case. The experiment did not prove AI juries work. It proved AI can mimic the deliberation process well enough to produce plausible-looking outcomes.

The problem that does not go away

The strongest counterargument to algorithmic adjudication is not philosophical - it is empirical. COMPAS, the AI risk assessment tool used in US criminal sentencing, was found by ProPublica to systematically classify Black defendants as higher risk than white defendants with comparable profiles. The algorithm's inner workings were, and remain, a trade secret. Courts used it anyway. The bias went largely unchallenged until researchers pulled the data.

That is the structural problem with putting AI in the decision seat: when the model is wrong, accountability dissolves. A human judge who makes a discriminatory ruling can be appealed, reversed, and ultimately removed. An algorithm that encodes discriminatory patterns at the data level produces outcomes that look procedurally clean while producing systematically unfair results. Harvard researchers studying LLM consistency for legal interpretation found that models fail what they called a textualist Turing test - they are not stable enough in their reasoning to be trusted as final arbiters.

There is also the hallucination problem. In early 2026 alone, a US appeals court ordered a lawyer to pay a $2,500 fine over AI-fabricated case citations in a brief. A district attorney was sanctioned for undisclosed AI use in court filings. Major law firms have apologized for submitting briefs riddled with AI-generated errors. If lawyers using AI as an assistant are still producing this volume of errors, the risk of AI as the decision-maker is considerably higher.

The asymmetry that makes the question urgent

The Hormozi scenario assumes symmetric AI use - both sides using the same tool, producing a kind of canceling-out that makes the human attorney layer redundant. But that is not the current reality. AI adoption in legal practice is uneven. Large firms with resources are deploying sophisticated legal AI. Public defenders and individuals representing themselves are not. In March 2025, a man without legal representation tried to use an AI avatar to argue his civil appeal in a New York court. Judges dismissed it immediately.

If AI makes elite legal argumentation cheaper for well-funded parties while access barriers remain for others, it does not flatten the playing field - it steepens it. The case for AI judges becomes most appealing in low-stakes, high-volume disputes where the current system is too slow and expensive to be practical. Some legal observers expect AI co-mediators and experimental AI court pilots for minor disputes to emerge in 2026. That is a different and narrower claim than replacing judges in criminal or complex civil proceedings.

What the transition actually looks like

The more credible near-term picture is hybrid rather than replacement. AI handles document review, legal research, precedent analysis, and administrative scheduling. Human judges retain final decision authority but lean increasingly on AI-generated summaries and risk scores. Juries are assisted by AI tools that clarify legal instructions and help evaluate evidence - addressing the documented problem that roughly 17% of jurors in studied cases did not understand the judge's instructions.

Kazakhstan has already moved in this direction, deploying AI inside its judicial system to analyze legislation and surface consistent precedents across similar cases - while legally enshrining that AI is a tool, not a decision-maker. The country's 2025 AI law explicitly anchors responsibility in humans. That framing - AI as infrastructure for better human decisions rather than a replacement for human judgment - is where most serious legal and technology institutions currently land.

The Hormozi framing is useful not because it describes the next decade, but because it describes the direction of pressure. Each efficiency gain at the attorney level makes the human layer look more like overhead. The question courts, regulators, and the public need to answer is not whether AI will assist in legal decisions - it already does - but where the accountability boundary sits when it gets something wrong.

Sources