The University of Chicago’s new policy is not an AI ban. It is an attempt to teach future lawyers a skill that every prisoner, family advocate and pro se litigant also needs: how to use artificial intelligence without surrendering judgment to it.
The University of Chicago Law School recently released one of the clearest institutional strategies yet for dealing with artificial intelligence in legal education.
Social media quickly reduced the announcement to a simpler headline: Chicago is taking laptops and AI away from first-year law students.
That description is incomplete.
Chicago is not trying to produce lawyers who never use AI. It is trying to produce lawyers who can still think when AI is unavailable—and who know enough law to recognize when AI is wrong.
That distinction matters.
It matters in law schools. It matters in law offices. And it may matter even more inside prisons and in the homes of families trying to challenge convictions, prison conditions and government decisions without meaningful legal assistance.
WHAT THE UNIVERSITY OF CHICAGO ACTUALLY ANNOUNCED
During the 2026–2027 academic year, the Law School will pilot a coordinated policy for its required first-year doctrinal courses.
Students generally will not be permitted to use laptops, phones or tablets in those classrooms. Limited exceptions will exist for disability accommodations, specific technology-based activities and designated classroom note-takers. Examinations in the core courses will be conducted in person without access to the internet, electronic files or apps.
But that is only one part of the strategy.
In first-year legal research and writing, students will work both without AI and with AI. The school says traditional writing will serve as the foundation, while students will also use AI for research, revision, developing drafts and preparing for oral argument. Faculty members and students will review both the writing and the student’s use of AI.
Upper-level professors will have greater flexibility. Clinics are acquiring legal AI tools and developing rules designed to prevent AI-created errors in court filings and other legal work.
Students completing the substantial research paper required for the JD degree will also have to discuss their work in person with a supervising professor. They may be questioned about the paper’s reasoning, conclusions and broader implications.
This is not an institution running away from artificial intelligence.
It is an institution trying to separate three different abilities:
1. Thinking without AI.
2. Working effectively with AI.
3. Knowing when AI should not be trusted.
That is a far more serious approach than simply declaring AI either forbidden or revolutionary.
YOU CANNOT SUPERVISE WORK YOU DO NOT UNDERSTAND
Chicago’s strongest argument is also the simplest:
A person cannot responsibly supervise AI-generated legal work unless that person understands the legal work.
Beginning law students have not yet developed the knowledge required to evaluate everything an AI system produces. A polished answer may look persuasive while misunderstanding the governing rule, overlooking a procedural problem or making an argument unsupported by the record.
The Law School says students must develop critical thinking and professional judgment rather than using AI shortcuts that produce easy answers while weakening intellectual growth. At the same time, the policy does not attempt to eliminate every useful application of AI. Students may still use it to clarify concepts, create practice questions and strengthen their engagement with the material.
That principle applies directly to incarcerated people, their families and pro se litigants.
AI can be extraordinarily useful for:
• Organizing a case timeline.
• Explaining unfamiliar legal language.
• Comparing reports, transcripts and witness statements.
• Locating apparent contradictions.
• Developing research questions.
• Improving the structure of a letter, grievance or motion.
• Identifying issues that may require investigation.
• Turning disorganized records into a workable factual summary.
But none of those uses makes AI a lawyer.
It has no professional license, no duty of loyalty, no personal knowledge of the case and no responsibility for the consequences of what gets filed.
The person using the output remains responsible for understanding it.
THE ORAL-DEFENSE REQUIREMENT MAY BE THE BEST PART OF THE POLICY
One LinkedIn commentator summarized the new research-paper requirement with a sharp line:
“AI can write a paper. It cannot defend one in real time.”
That statement is not literally accurate. An AI system can generate real-time answers when someone has access to it.
But the underlying point is correct.
The student must be able to defend the paper without leaning on the machine.
Chicago’s new requirement is designed to test whether students understand what they submitted. Professors will ask questions that examine the paper’s reasoning and the consequences of its argument. The school says this mirrors the reality of legal work, where lawyers must explain and defend their positions in courtrooms, negotiations, client meetings and discussions with other lawyers.
The Prison Project believes every AI-assisted legal document should survive a similar test.
Before submitting a grievance, administrative appeal, habeas petition, motion or court brief, the person filing it should be able to answer:
• What happened?
• What exact decision or action is being challenged?
• Which facts support the claim?
• Where are those facts documented?
• What legal rule supposedly applies?
• Has that rule been verified in an official or reliable source?
• What deadline controls?
• What procedural requirements must be followed?
• What relief is being requested?
• Why does the court or agency have authority to grant that relief?
If the person cannot explain the document paragraph by paragraph, it is not ready to be filed.
That does not mean the person must have the education of a law professor. It means they must understand the argument being made in their name.
THE HIDDEN INEQUALITY IN CHICAGO’S POLICY
The University of Chicago can build a carefully controlled system around AI because its students have professors, libraries, clinics, feedback, legal databases and opportunities to correct mistakes.
Incarcerated people often have none of those advantages.
The United States Supreme Court has held that the federal constitutional right to appointed counsel generally does not extend beyond the first appeal as of right and does not ordinarily require the state to provide a lawyer for collateral post-conviction proceedings.
That is the stage at which many prisoners must raise claims involving ineffective assistance of counsel, newly discovered evidence, suppressed evidence, constitutional violations or other challenges to a conviction that has already become final.
The rules are complicated. Deadlines can be unforgiving. Procedural default can prevent a court from ever reaching the underlying claim.
Yet the person expected to navigate that system may be working from a prison law library with restricted hours, missing materials and limited communication with family members or potential counsel.
Even where a person does have a lawyer, incarceration can interfere with meaningful access. A Justice Department review of federal pretrial facilities documented delays and barriers involving legal visits, telephone calls and review of discovery. That report was limited to federal pretrial detention, but it demonstrates a broader reality: the existence of a lawyer on paper does not always mean the incarcerated client can communicate with that lawyer effectively.
This is where the discussion about “human lawyering” becomes uncomfortable.
Chicago correctly says that clients, judges, employers and society will continue to want humans to perform critical legal functions such as advocacy, strategic judgment, critical thinking and building relationships with clients.
Of course people want human lawyers.
The real question is: Who can obtain one?
There is little comfort in being told that human judgment is essential when the legal system does not provide a human lawyer and the incarcerated person cannot afford to hire one.
For many prisoners and families, AI is not replacing a lawyer.
There was no lawyer available to replace.
BANNING AI WOULD PRESERVE THE INEQUALITY
The answer is not to hand incarcerated people an unrestricted chatbot and tell them to file whatever it produces.
That would be reckless.
But completely denying prisoners access to responsible AI assistance would also be reckless. It would reserve one of the most powerful educational and organizational tools ever developed for lawyers, law firms, universities and people with money.
The same institutions warning that prisoners might misuse AI are increasingly using it themselves.
Law schools are training students to work with it. Clinics are obtaining specialized AI products. Employers expect graduates to understand it. Chicago itself recognizes that AI is already embedded in legal practice and that it is unrealistic to pretend lawyers will not use it.
A two-tier system would be indefensible:
• Lawyers use AI to review documents, organize evidence and improve research.
• Prisoners are forbidden from using AI to understand the documents controlling their own lives.
The proper answer is not prohibition.
It is structured access, education, verification and accountability.
THE PRISON PROJECT STANDARD FOR RESPONSIBLE LEGAL AI
The Prison Project does not treat AI as a lawyer, a court or an authority.
We treat it as a tool.
A powerful tool can help people understand a case. It can also help them produce confident nonsense more quickly. The difference depends on how it is used.
Our working standard is:
1. USE AI TO IDENTIFY QUESTIONS—NOT DECLARE FINAL ANSWERS
AI may help identify a possible issue. That issue must then be researched and verified.
A chatbot saying that a constitutional violation occurred does not establish that a constitutional violation occurred.
2. VERIFY EVERY LEGAL AUTHORITY
Every case, statute, regulation and court rule must be checked against a reliable source.
Confirm that the authority exists, says what the AI claims it says, remains good law and applies in the correct jurisdiction.
3. TIE EVERY FACTUAL CLAIM TO THE RECORD
AI should never be permitted to fill gaps in the evidence.
Every material factual statement should be connected to a transcript page, report, affidavit, medical record, grievance, letter, photograph or other identifiable source.
4. SEPARATE EVIDENCE FROM ALLEGATION
A family member’s belief may justify investigation, but it is not automatically proof.
AI-generated writing often makes allegations sound more certain than the underlying evidence supports. The language must accurately reflect what is known, what is disputed and what remains unverified.
5. CONFIRM DEADLINES AND PROCEDURE INDEPENDENTLY
A strong constitutional argument can still fail if it is filed in the wrong court, under the wrong procedure or after the deadline.
AI assistance does not excuse procedural mistakes.
6. DO NOT SUBMIT ANYTHING YOU CANNOT EXPLAIN
The person signing or submitting the document should understand its purpose, factual foundation, legal theory and requested relief.
The final question should always be:
Could I defend this document if a judge asked me about it without giving me access to the chatbot?
If the answer is no, more work is required.
HUMAN JUDGMENT IS NOT OBSOLETE
The University of Chicago is right about something deeper than classroom technology.
Legal work is not simply the production of words.
It involves judgment.
It requires deciding which facts matter, which witnesses are credible, which arguments are worth making, which risks are acceptable and what a frightened client needs to understand before making a life-changing decision.
AI can assist with those tasks. It cannot assume the moral and professional responsibility that comes with them.
But praising human judgment is not enough.
Institutions must also confront whether ordinary people—especially incarcerated people and their families—have access to anyone capable of exercising that judgment on their behalf.
When they do not, responsible AI access may help narrow the gap.
It will not replace counsel. It will not cure an unjust system. It will not transform a weak claim into a winning one.
But it can help a person understand the record, ask better questions, communicate more clearly and recognize when professional assistance is urgently needed.
That is not a threat to legal education.
That is legal education reaching people who have historically been denied it.
THE REAL LESSON
Chicago Law’s policy should not be remembered as the moment a law school banned laptops.
Its more important lesson is this:
Learn to think without AI so that you can think more effectively with it.
For prisoners and families, we would add one more line:
Use AI to strengthen your ability to question the system—never to surrender your judgment to another system you do not understand.
Artificial intelligence should not become another gatekeeper between ordinary people and the law.
Used carefully, it can become a tool for opening that gate.
SOURCES
University of Chicago Law School, “AI Strategy Statement”:
https://www.law.uchicago.edu/news/ai-strategy-statement
University of Chicago Law School, “UChicago Law Unveils New AI Strategy Statement”:
https://www.law.uchicago.edu/news/uchicago-law-unveils-new-ai-strategy-statement
Pennsylvania v. Finley, 481 U.S. 551 (1987):
https://supreme.justia.com/cases/federal/us/481/551/
United States Department of Justice, Access to Counsel in Federal Pretrial Detention:
https://www.justice.gov/d9/2023-07/2023.07.20_atj_bop_access_to_counsel_report.pdf
PrisonProject.net provides legal education, case-analysis tools and advocacy resources for incarcerated people and their families. Our materials are educational and are not a substitute for advice from a licensed attorney.
Comments