1. Introduction

Artificial intelligence (AI) courtroom decision support governance in Morocco concerns the legal rules, institutional safeguards, and accountability mechanisms required when AI tools assist judges, court clerks, lawyers, and judicial administrators in resolving civil disputes.

AI courtroom systems may help courts organize evidence, summarize documents, identify relevant legislation, retrieve precedents, estimate damages, manage case files, and identify procedural delays. However, their use raises fundamental questions about judicial independence, equality before the law, privacy, transparency, the right to be heard, and the ability to challenge an incorrect decision.

The central principle is that AI may assist Moroccan judges, but judicial responsibility must remain with the legally competent human decision-maker. An algorithm should not independently determine civil rights, resolve disputed facts, or issue binding judgments without a valid legal basis and appropriate safeguards.

Morocco's framework must be understood through its constitutional principles, civil procedure, obligations and contracts law, personal data protection legislation, and applicable international human-rights commitments. The governance of AI in courts should build on these existing legal foundations rather than assume that a comprehensive, AI-specific judicial code is already in force.

2. Legal framework governing AI courtroom decision support in Morocco

A. Constitution and judicial independence

Morocco's Constitution of 2011 establishes the institutional and rights-based foundation for judicial decision-making.

The relevant principles include:

Judicial independence: judicial decisions must be made independently of improper executive, commercial, or technological influence.

Equality before the law: an AI system must not systematically disadvantage litigants because of language, gender, disability, income, geographic location, or other irrelevant characteristics.

Fair trial and defence rights: parties must have a meaningful opportunity to present evidence, challenge opposing evidence, and respond to material that may affect the outcome.

Reasoned judicial decisions: a judge must be able to explain the legal and factual grounds for a judgment rather than simply adopt an algorithmic recommendation.

An AI vendor should not be permitted to determine the substantive outcome of a dispute through undisclosed scoring rules or unreviewable recommendations.

B. Law No. 09-08 on personal data protection

Morocco's Law No. 09-08, promulgated in 2009, is especially important where courtroom AI processes personal data. The National Commission for the Control of Personal Data Protection (CNDP) oversees the relevant data-protection framework.

Chambers and Partners

+1

Key requirements and implications include:

Lawful processing: information must be processed on a valid legal basis and for legitimate, specified purposes.

Data minimisation: systems should not collect or retain more personal information than necessary.

Accuracy: inaccurate names, financial records, case histories, or extracted facts must be capable of correction.

Information and access rights: applicable statutory rights must be respected, subject to lawful restrictions.

Automated decisions: Article 11 restricts judicial decisions based on automated processing used to assess a person's conduct or personality. It must be read carefully rather than treated as a blanket prohibition on every use of AI in court administration.

LexLint

+1

For example, an AI tool may help a judge search previous property judgments. That is materially different from a system that assigns a litigant a risk score and automatically determines whether that person should win a property dispute.

C. Cybersecurity and confidentiality

Law No. 05-20 on cybersecurity is also relevant to covered public entities and critical infrastructure. Court systems must consider unauthorised access, data leakage, manipulation of evidence, and the integrity of digital records.

Chambers and Partners

Sensitive court information may include:

Family disputes and child-protection records.

Medical reports and disability information.

Financial statements and banking records.

Commercial secrets and confidential contracts.

Witness statements and protected personal information.

AI systems should use access controls, encryption, secure logging, appropriate retention periods, and incident-response procedures proportionate to the sensitivity of the information.

D. Current position on AI-specific court rules

As of October 2026, available legal-regulation research indicates that Morocco has not adopted a comprehensive, binding AI-specific judicial decision-support regime. The CNDP has addressed AI-related data protection, while discussions involving the judicial authorities and UNESCO have examined safeguards for AI in courts.

SafeLegalAI

+1

Consequently, the governance model should distinguish between existing legal obligations and proposed safeguards. A court cannot assume that a technology is lawful merely because it improves administrative efficiency.

3. Practical uses of AI in Moroccan civil courts

1. Document classification and case management

AI can sort pleadings, identify missing documents, extract dates, and flag filing deadlines. Staff must verify classifications before procedural consequences follow.

2. Legal research and precedent retrieval

Systems can locate relevant statutes and earlier judgments. Every cited authority must be checked for authenticity, legal status, and relevance to the dispute.

3. Hearing transcription and translation

Speech recognition may assist with Arabic, French, and other languages used in proceedings. Human review is essential for names, legal terminology, dialects, and disputed statements.

4. Outcome and damages analysis

Statistical tools can compare historical outcomes or assist with calculations. Historical patterns must not be mistaken for legal rules, and a prediction cannot replace an assessment of the evidence.

Morocco has already explored AI-assisted legal-document processing. A 2025 research study described a system for classifying Moroccan criminal judgments and extracting information from them. Earlier studies examined machine-learning prediction of accident-case outcomes and prosecutorial decisions. These projects demonstrate technical possibilities, not proof that automated judicial determinations are legally appropriate.

ScienceDirect

+2

4. Six principal governance challenges

1. Algorithmic bias and discrimination

AI systems learn from historical records. If older judgments reflect inconsistent practices or unequal treatment, an algorithm may reproduce those patterns.

For example, a system trained on past compensation awards might systematically recommend lower damages for certain categories of claimants because of historical disparities rather than legally relevant evidence.

Governance response: conduct bias testing, evaluate outcomes across relevant groups, document training-data limitations, and permit judges to reject recommendations with recorded reasons.

2. Transparency and explainability

A judge and the parties should be able to understand the relevant basis of an AI-generated recommendation. A proprietary system should not make it impossible to identify a factual error, a misapplied rule, or a discriminatory assumption.

A useful explanation should identify the information relied upon, the limitations of the analysis, and the distinction between established law and statistical prediction.

3. Human judicial control

A judge must independently evaluate the evidence and law. Merely approving a computer-generated result without genuine scrutiny creates a risk of automation bias.

Human oversight must therefore be substantive: the judge should have sufficient expertise, access to the underlying record, and authority to disregard the system's output.

4. The right to challenge AI-assisted evidence

If an AI-generated analysis materially influences a civil judgment, the parties should have an effective opportunity to challenge its reliability, methodology, and relevance, subject to the applicable procedural rules.

A court should not rely on a decisive undisclosed assessment while denying the affected party any meaningful opportunity to respond.

5. Privacy and confidentiality

Court databases may contain sensitive personal, financial, medical, and family information. Reusing those records to train general-purpose AI models creates additional risks.

Governance measures should include purpose limitation, appropriate anonymisation, restricted access, vendor confidentiality obligations, and controls on the transfer or reuse of court data.

6. Accountability for errors

Responsibility must be allocated clearly among the judicial institution, system provider, court staff, and other relevant actors.

A provider's contractual disclaimer should not remove a litigant's lawful remedies. Where an AI error contributes to an unlawful judgment, the applicable rules on appeal, procedural correction, liability, and data protection must remain available.

5. Case law: at least six relevant judicial decisions

Important legal qualification: I could not verify six Moroccan judgments directly deciding the legality of AI courtroom decision-support systems. The following cases are therefore divided into one relevant Moroccan decision concerning technical expertise and six comparative authorities. The comparative cases are useful for developing safeguards, but they are not binding Moroccan precedents. They should not be cited as if Moroccan courts have already adopted their holdings on AI.

Case 1. Moroccan Court of Cassation — Decision No. 5/7, 13 January 2000

Reference: Criminal file No. 17203/6/7/99

The reported decision concerned judicial treatment of expert evidence relating to a person's mental condition. The reported principle emphasises that a technically specialised question should be addressed through appropriate expertise and that a court should not casually disregard properly obtained expert findings.

Relevance to AI: an algorithm's output should not be treated as self-validating technical truth. Its reliability, methodology, limitations, and relationship to the evidence must be examined. The case does not itself establish a rule on AI and must be applied cautiously because it concerned criminal proceedings and expert evidence.

cnej.ma

Case 2. Taxquet v. Belgium (European Court of Human Rights, Grand Chamber, 2010)

The European Court of Human Rights examined whether a criminal verdict provided adequate reasons to enable the accused to understand why they had been convicted.

Principle: judicial outcomes must provide sufficient safeguards for understanding the basis of the decision.

Relevance to AI: if an AI system materially influences a Moroccan civil judgment, the judge's reasons should explain the actual factual and legal grounds for the outcome. A statement that the algorithm recommended a result would not, by itself, provide adequate judicial reasoning.

This is a comparative fair-trial authority, not a Moroccan AI ruling.

Case 3. Ruiz Torija v. Spain (European Court of Human Rights, 1994)

This case concerned the obligation to give reasons for judicial decisions and the failure to address a significant argument raised by a litigant.

Principle: courts must adequately address material submissions; they are not necessarily required to respond individually to every minor argument.

Relevance to AI: an AI summary must not cause the judge to overlook a decisive submission, defence, or piece of evidence. Judges should compare automated summaries with the original pleadings and ensure that the judgment addresses the issues essential to the dispute.

Case 4. Dombo Beheer B.V. v. the Netherlands (European Court of Human Rights, 1993)

The Court considered equality of arms in civil proceedings, including an imbalance in the parties' opportunities to present evidence.

Principle: each party must have a reasonable opportunity to present its case without being placed at a substantial disadvantage relative to the opposing party.

Relevance to AI: if one party has access to sophisticated AI-generated analysis while the other cannot meaningfully inspect or challenge it, procedural inequality may arise. Moroccan courts should consider disclosure, adequate response time, and equal opportunities to challenge material algorithmic evidence.

Case 5. S. and Marper v. the United Kingdom (European Court of Human Rights, Grand Chamber, 2008)

The case concerned the retention of fingerprints, cellular samples, and DNA profiles by public authorities and the protection of private life.

Principle: the collection and retention of personal information by public authorities must respect privacy rights and applicable legal safeguards.

Relevance to AI: court AI systems should not retain identifiable litigant data indefinitely or reuse it for unrelated purposes without a lawful basis. The case supports privacy-oriented safeguards but does not directly determine the legality of judicial AI.

Case 6. R (Bridges) v. Chief Constable of South Wales Police (England and Wales Court of Appeal, 2020)

The case concerned the police use of automated facial-recognition technology in public spaces.

Principle: public authorities using algorithmic technologies must act within their legal powers and comply with applicable privacy, equality, and public-law obligations.

Relevance to AI: Moroccan judicial institutions should establish a clear legal basis for high-impact AI uses, assess discriminatory effects, and document how the system is deployed. Public institutions cannot assume that a technology is lawful simply because it is technically effective.

Case 7. State v. Loomis (Wisconsin Supreme Court, United States, 2016)

The case concerned the use of the COMPAS risk-assessment tool during criminal sentencing.

Principle: the court considered the permissible use and limitations of a proprietary algorithmic risk assessment, including concerns about transparency and the possibility of overreliance.

Relevance to AI: judicial decision-support tools should not be treated as definitive assessments of a person's credibility, conduct, or likely behaviour. Limitations, potential bias, and the system's proper role must be understood, and the judge must retain independent responsibility.

This was a criminal sentencing case, not a civil dispute or Moroccan judgment.

Case 8. State of Maharashtra v. Dr. Praful B. Desai (Supreme Court of India, 2003)

The Court considered the use of video conferencing in the taking of evidence.

Principle: judicial procedure can accommodate technology where its use is consistent with the governing law and the relevant procedural safeguards.

Relevance to AI: digitisation, remote participation, and AI-assisted administration need not be rejected categorically. Instead, courts should assess whether the particular technology preserves procedural fairness, reliable evidence, and meaningful participation.

This is a comparative Indian authority concerning video conferencing, not an AI decision-support ruling.

How these cases fit together

These authorities provide a comparative analytical framework. Their legal weight in Morocco differs: Moroccan legislation and relevant Moroccan judicial decisions govern domestic disputes, while foreign decisions can offer persuasive reasoning where appropriate.

6. Proposed governance framework for Moroccan courts

A practical governance framework should contain the following safeguards.

1. Judicial authority

AI recommendations must remain advisory. Judges retain responsibility for findings of fact, interpretation of law, and the final judgment.

2. Mandatory risk assessment

High-impact systems should undergo legal, technical, privacy, cybersecurity, and discrimination assessments before deployment.

3. Traceability and disclosure

Maintain records of system versions, inputs, outputs, known limitations, human overrides, and material errors. Establish disclosure procedures when AI output may affect a party's rights.

4. Data protection

Apply Law No. 09-08, appropriate security controls, restricted access, and lawful retention and deletion policies.

5. Judicial training

Train judges, lawyers, and court personnel to identify hallucinated citations, biased recommendations, misleading statistics, and errors in Arabic- and French-language processing.

6. Independent audits and remedies

Periodically evaluate accuracy, disparate impacts, security, and complaints. Preserve access to ordinary judicial remedies and any applicable data-protection complaint mechanisms.

7. Illustrative application: an AI-assisted property dispute

Suppose a Moroccan court is deciding a dispute over ownership of land. An AI system reviews contracts, registration documents, prior judgments, and the parties' submissions.

A lawful and responsible workflow would be:

The court checks the legal basis for processing the parties' personal information.

The AI extracts relevant dates, identifies potentially applicable legal provisions, and retrieves comparable judgments.

Court personnel verify document extraction, while the judge checks the authenticity and legal relevance of every cited authority.

The parties are given an appropriate opportunity to challenge any AI-generated material that could materially affect the dispute.

The judge independently evaluates ownership documents, applicable property law, and the parties' arguments.

The final judgment explains the court's own reasoning and remains subject to the remedies available under Moroccan law.

The AI system must not decide ownership solely by matching the facts to historical outcomes. A previous judgment may concern different documents, legal provisions, or factual circumstances.

8. Recommendations for future Moroccan legislation

Morocco could strengthen its framework through a dedicated judicial AI instrument addressing:

Permitted and prohibited uses: distinguish administrative assistance from tools that materially influence judicial determinations.

Risk classification: apply stricter controls to systems affecting family rights, property, compensation, access to justice, and other important civil interests.

Explanation and contestability: establish procedures for challenging materially influential AI outputs.

Vendor accountability: require documentation, testing, security, audit access, and clear responsibility for system failures.

Judicial independence: prevent vendors, government departments, and external actors from improperly influencing individual judgments.

Public oversight: publish appropriate information about approved systems, audits, incidents, and performance without disclosing protected case information.

Periodic review: reassess systems as models, data, legal rules, and risks change.

These are recommendations for future reform, not a statement that Morocco has already enacted all these requirements.

9. Conclusion

AI courtroom decision-support governance in Morocco lies at the intersection of civil procedure, constitutional safeguards, personal-data protection, cybersecurity, and judicial independence. AI can help reduce administrative burdens, improve legal research, and organise complex records, but its usefulness does not establish the legality of every application.

The most defensible approach combines human judicial control, lawful data processing, technical reliability, procedural equality, meaningful explanations, and effective remedies. Morocco's Law No. 09-08 provides an important existing safeguard concerning certain automated judicial decisions, while comparative decisions such as Bridges, State v. Loomis, and Dombo Beheer offer useful lessons about algorithmic accountability, judicial overreliance, and fair proceedings.

Ultimately, the legitimacy of a Moroccan civil judgment must rest on applicable law, verified evidence, fair procedure, and the independent reasoning of the competent judge—not on the authority of an AI system.

Educational note: This analysis is for legal research and academic study. The applicability of individual statutory provisions and cases should be verified against authoritative Moroccan legal texts before use in litigation or formal legal advice.