Guidelines by field
AI guidelines in law and the courts what the field's own bodies ask
Law is the field where unchecked AI output meets immediate penalties. Since 2023 courts have fined and referred lawyers who filed invented cases, and bar associations and judges have written guidance on competence, confidentiality and responsibility.
Last reviewed 27 October 2026. Every claim links to its source.
At a glance
Competence
Understand enough about the tool, including its limits, to use it responsibly.
Confidentiality
Do not put client information into a tool unless you know how it stores and reuses data. UK judges are told to treat anything typed into a public chatbot as potentially public.
Verify every citation
The High Court in London said generative AI tools are not capable of reliable legal research.
You answer for what bears your name
Whether you or a junior used the AI.
Judges keep the judging
Core judicial decisions are not handed to AI.
Who has spoken
- Binding law, regulation, court ruling or enforceable professional rule
- Voluntary guidance, principles or a code an organization chooses to follow
Voluntary Jul 2024
American Bar Association
Formal Opinion 512: Generative AI Tools
Applies existing rules to AI: competence, confidentiality and informed consent, telling clients when it matters, supervising others, and billing for real time spent rather than for learning the tool. Advisory, but courts and disciplinary bodies rely on it.
Voluntary Nov 2023 to Apr 2024
State bars of California, Florida and New York
Practical guidance and ethics opinions
State-level guidance on confidentiality, supervision and fees, interpreting the enforceable rules of professional conduct.
Binding Jun 2025
High Court of England and Wales
Ayinde v Haringey; Al-Haroun v Qatar National Bank
Warned that knowingly placing false material before a court could amount to contempt or a criminal offence, and referred lawyers to their regulators.
Voluntary Oct 2025
Courts and Tribunals Judiciary (England and Wales)
AI guidance for judicial office holders
Updated twice in 2025. Covers confidentiality, bias, hidden "white text" prompts and deepfaked evidence.
Voluntary Jul 2025
US federal judiciary
Interim AI guidance for courts
Allows experimentation with guardrails, keeps judges responsible, and bars delegating core judicial decisions to AI.
Voluntary 2025 to 2026
US Judicial Conference rules committees
Proposed Federal Rule of Evidence 707
A proposed rule on machine-generated evidence went out for comment and was sent back for more study in June 2026. Not adopted.
Not capable of conducting reliable legal research.
Binding and voluntary
Binding
The rules of professional conduct, enforced by bar discipline; court sanctions, wasted-costs orders and referrals to regulators.
Voluntary
Bar opinions interpret the binding rules. Judicial guidance in the UK and the US federal interim guidance govern judges' own conduct but are not statutes.
What changed in 2025 and 2026
- Hallucinated citations became a routine disciplinary problem: one researcher's database counted 1,240 reported US instances from April 2023 to July 2026, 668 of them in 2026, and the most common fine was $1,000 to $5,000.
- The UK judiciary updated its guidance twice in 2025, and US federal courts issued interim guidance in July 2025.
- A proposed US rule on machine-generated evidence was sent back for more work in June 2026.
Where people still disagree
How courts should handle evidence made or altered by machines is still open; the proposed US evidence rule was contested and returned for study.
Questions to ask yourself
- Have I personally read every case and source cited in this document?
- Would my client be comfortable knowing which of their details went into this tool?
- If the judge asked "how did you use AI here?", could I answer completely?
How this lines up with my everyday guidelines
| My everyday guidelines | The field's nearest rule |
|---|---|
2. Practice transparent collaboration Own all outputs that carry your name or business reputation. | You remain responsible for anything filed in your name (ABA 512; Ayinde). |
6. Stay curious and critical Question outputs. AI can be confident and wrong. | Verify every citation and fact against authoritative sources (High Court). |
5. Protect privacy and boundaries Be mindful of confidential business or family information. | Do not put client information into a tool without knowing how it reuses data (ABA 512). |