Hollywood is already asking what happens when an AI decides who goes free and who does not. In the 2026 thriller Mercy, Chris Pratt stars in a near-future story in which an AI-administered justice system decides a murder case. It is a dramatic "what if" — and it raises a question worth taking seriously: how close is real life to that?
The answer, based on what the federal judiciary is actually doing right now, is: not close at all. But real courts are moving carefully with AI, and the reasons for their caution matter to anyone who uses AI for serious work.
What Is Actually Happening
The US federal judiciary is preparing new guidance on the use of AI in the courts. According to reporting by Reuters summarized in a PYMNTS CPI post, new recommendations could be developed within the next year. The effort is led by US Circuit Judge Sidney Thomas, who chairs a federal judiciary task force on generative AI.
The task force is a 24-member group established in 2025. It has identified more than 60 areas requiring examination, covering cybersecurity, governance, ethics training, probation and pretrial operations, federal defender services, and the technology's costs. So far it has produced about 30 recommendations.
The central tension is straightforward. Courts see AI's potential to improve efficiency, but they weigh that against errors, security risks, and the danger of delegating judicial responsibilities that belong to human judges.
The Hard Line: Judges Keep the Judgment
One concern stands above the others: whether judges could become overly dependent on generative AI for duties that require independent judicial judgment.
The judiciary has already drawn a line. Guidance distributed to federal judges in July 2025 told them not to hand core judicial functions to AI systems and to independently check anything the technology produced. That guidance was not released publicly, but its contents were described in an October letter from the judiciary to Sen. Chuck Grassley, chairman of the Senate Judiciary Committee.
The letter followed real incidents. Two judges acknowledged using AI in preparing court orders that turned out to contain mistakes. Courts have also confronted fabricated case citations and other inaccurate AI-generated material in legal filings, which has sometimes led to sanctions against attorneys and people representing themselves.
Where Courts Are Actually Using AI
The uses courts are exploring today are administrative, not adjudicative.
Judge Thomas pointed to the San Francisco-based 9th US Circuit Court of Appeals, where 49% of appeals in the previous year involved people representing themselves. Some of those litigants submit handwritten documents. AI has been used to convert those filings into more readable text, reducing work for court employees and making submissions easier to review.
The 9th Circuit, along with the 5th and 6th Circuits, has also deployed AI as a quality-control tool for the large volume of briefs courts receive. The systems flag filings that fail procedural or formatting requirements. US Circuit Judge Jeffrey Sutton, who chairs the Judicial Conference's executive committee, said the screening lets court clerks alert litigants to defects so they can submit corrected documents rather than risk having an appeal dismissed for technical reasons.
That is the pattern worth noticing: AI reading, sorting, flagging, and preparing — with a human deciding.
How Close Is Real Life to Mercy?
Not close. Nothing in the current effort points toward AI judges deciding cases.
In Mercy, the premise is that an AI system administers justice: it weighs a murder case and renders a decision. In the real federal courts, the current direction is the opposite. Keep AI out of the core act of judging. Use it to reduce paperwork and catch formatting errors. Require humans to verify anything AI produces.
The honest answer to this article's title is no. Real-life Mercy-like judges are not coming soon. What is coming is something quieter: careful rules about where AI may help and where it may not, written after mistakes had already happened.
Why This Matters Beyond the Courtroom
The federal judiciary is working through the same governance problem every ordinary person and small-business owner faces when AI gets involved in serious work: AI is fast, but it can be confidently wrong.
The principles courts are settling on translate directly to everyday work:
- Keep the human accountable for the decision.
- Let AI prepare, flag, and summarize — but not decide on its own.
- Check AI output before it affects someone else.
- Keep a record of what the AI produced and what the human changed.
These are the same disciplines that make AI useful rather than risky in a business. If you use AI to draft customer messages, invoices, applications, or anything with legal or financial weight, treat the output the way federal judges are now required to treat it: as a first draft, not a final answer.
A Practical Takeaway
You do not need to follow courtroom policy to adopt the same habits. Two small tests cover most of it:
- For anything consequential, write down the one thing AI must not decide for you. For a judge, that is the judgment itself. For you, it might be pricing, a commitment to a customer, or a filing that carries legal weight.
- Before any AI-prepared material leaves your desk, check the facts it contains — names, dates, numbers, citations, and requirements — the way the judiciary now requires judges to check AI output.
Tools built for this kind of work make the habit easier. Bot Commander applies the same governance principle to business workflows: consequential actions such as sending, publishing, spending, and account changes stop for owner approval, and the prepared work stays visible for review. It is the same idea the federal courts are formalizing — AI handles the preparation, the human keeps the decision.
If you want to track where this is heading, watch what the judiciary's task force recommends over the next year. The details of federal court policy will matter less to most readers than the pattern they reveal: the institutions we trust most are keeping humans in the loop.
---
Adapted from PYMNTS's "US Judiciary Prepares New AI Guidance for Federal Courts" — https://www.pymnts.com/cpi-posts/us-judiciary-prepares-new-ai-guidance-for-federal-courts/