Update: UNESCO Just Drew the Red Lines for AI in Courts – Here’s What They Really Mean From “AI as assistant” to “AI as decision-maker”: how UNESCO is quietly reshaping the future of justice.

UNESCO Just Drew the Red Lines for AI in Courts – Here’s What They Really Mean

From “AI as assistant” to “AI as decision-maker”: how UNESCO is quietly reshaping the future of justice.

---

Introduction – AI Is Already in Courtrooms

Artificial intelligence is no longer a “future trend” in justice systems.

Courts around the world already use AI for:

• Legal research over massive databases
• Case management and workload allocation
• Translation and transcription during hearings
• Document classification and anonymisation of judgments
• Drafting support and summarising long decisions

In other words: AI is now woven into the everyday machinery of justice.

But courts are not like other institutions.

A judicial decision can affect:

• Liberty – who is detained, who is free
• Property – what people keep, what they lose
• Family – custody, guardianship, personal status
• Reputation & Dignity – how someone is seen by society

That makes one question unavoidable:

Not whether courts should use AI – but how they can use it without breaking justice itself.

This is exactly the question UNESCO tried to answer in 2025.

---

Why UNESCO Stepped In

In 2025, UNESCO issued the Guidelines for the Use of AI Systems in Courts and Tribunals – the first global ethical and operational framework designed specifically for judicial AI.

A few things make this document unique:

• It is sector-specific: only about courts and tribunals, not generic “AI ethics”.
• It is human-rights based: grounded in fair trial guarantees, equality, dignity, and the rule of law.
• It is globally informed: developed after consultations with more than 36,000 judicial actors in over 160 countries.
• It is already being cited and used by courts, including constitutional courts.

The message is very clear:

AI may assist justice – it must never replace judicial reasoning or human judgment.

UNESCO is not trying to stop innovation. It is trying to stop unregulated automation of power.

---

Assistant vs. Decision-Maker – The Red Line in Judicial AI

The single most important distinction in the Guidelines is this:

• AI as an assistant vs.
• AI as a decision-maker

Everything else builds on this difference.

AI as an assistant

Here, AI:

• searches huge volumes of case law and documents
• organises and classifies files
• translates hearings and documents
• summarises long texts
• supports court administration and scheduling

The judge remains fully in control:

• The judge understands how AI was used.
• The judge can accept, adapt, or reject AI outputs.
• AI is a tool, not an authority.

This “assistive AI” is where most current systems sit – and where UNESCO sees legitimate opportunity.

AI as a decision-maker

AI crosses the line when it:

• predicts or recommends outcomes
• weighs evidence and suggests how to interpret it
• proposes sentences or sanctions
• applies legal rules to facts in a way that shapes the judgment itself
• operates as a black box that even judges cannot explain

Even if a human signs the judgment at the end, the real reasoning may already have been pushed or constrained by the system.

UNESCO’s position is firm:

Any AI that replaces or silently steers judicial reasoning is incompatible with the rule of law.

Why? Because justice requires that decisions be:

• Reasoned – with a clear chain of logic
• Explainable – understandable to parties and appellate courts
• Challengeable – open to review and appeal
• Accountable – linked to a responsible human decision-maker

An algorithm cannot be cross-examined. It cannot be held morally or legally responsible. It does not “understand” justice.

---

The Human-Rights Foundation: Three Core Principles

Behind the detailed text, three legal pillars anchor UNESCO’s framework:

1. Human Rights as a Condition, Not a Decoration

Any AI system used in courts must respect:

• the right to a fair trial
• the presumption of innocence
• the right to defense and adversarial process
• privacy and data protection
• human dignity

If AI:

• is too opaque to be understood,
• prevents parties from knowing how it influenced proceedings, or
• blocks their ability to challenge its role,

?? it should not be used, regardless of speed or accuracy.

2. Non-Discrimination and Equality Before the Law

AI learns from data. Data often carries historical bias.

UNESCO therefore emphasizes:

• preventing algorithmic discrimination
• avoiding tools that reinforce or deepen existing inequalities
• safeguarding vulnerable and marginalized groups
• ensuring equal access to justice

If an AI system systematically produces worse outcomes for certain groups, or treats similar cases differently without justification, it is incompatible with the principle of equality before the law.

3. Proportionality: Not Every Problem Needs AI

Proportionality means:

• using AI only for a legitimate judicial purpose
• limiting its role to what is strictly necessary
• calibrating its use to the gravity of the legal impact

Examples:

• AI for document management or scheduling ? generally acceptable
• AI that influences sentencing, liberty, or core findings of fact ? extremely sensitive and often unacceptable

In short:

The more an AI system touches the substance of justice, the higher the bar for its use.