Why Judicial Officers Should Prepare
By Chisom Obiudo
A rejected loan application may seem an unlikely starting point for a court dispute about artificial intelligence (AI). Yet AI’s role in that decision could become central to a borrower’s complaint.
Picture this hypothetical scenario: a Namibian financial institution uses AI to assess loan applications from micro, small and medium-sized businesses.
The system was trained on historical lending decisions to predict which applications to approve.
Those records include repeated rejections of applicants from one region, even though their finances are comparable to those of applicants approved elsewhere.
The system learns that regional pattern and recommends similar rejections. AI is reproducing bias in past decisions.
This could raise concerns about unlawful discrimination. A court would need to examine why the applicants were treated differently and whether that treatment breached Namibian law.
Foreign cases show that disputes involving AI could reach our courts through familiar claims, such as those concerning loans, employment or misleading advice.
They could involve an individual professional, a small business, a public institution or a company. Judges and magistrates may need to examine AI-generated decisions, even if their courts never use AI.
From an AI decision to a legal claim
Such a claim requires a recognised legal basis, often called a cause of action. A contract claim asks whether an agreement was broken.
A discrimination claim examines why people were treated differently and whether the law prohibits that treatment.
The person bringing the claim normally has to prove the facts needed to obtain compensation or another remedy, such as an order to stop unlawful conduct.
Negligence requires further explanation. The question is whether someone failed to take precautions that a reasonably careful person in the same situation would have taken to prevent a foreseeable risk of harm.
The court considers what could reasonably have been expected at the time, rather than judging conduct in hindsight.
For professional work, the standard reflects the care and competence reasonably expected of that profession.
Imagine a lawyer accepting an AI-generated answer about a filing deadline without verifying the legislation or court rules.
The answer is incorrect, the deadline passes, and the client loses the opportunity to pursue a claim.
The court would assess whether a reasonably competent lawyer would have checked the deadline and whether that failure caused the client’s loss. The missed check matters; an AI mistake alone does not establish liability.
Lessons from foreign cases
These questions have already arisen abroad. In Moffatt v Air Canada (2024), a passenger asked the airline’s website chatbot about bereavement fares, which offer reduced fares for travel following a relative’s death.
The chatbot incorrectly stated that he could claim the discount after buying his ticket. He relied on that information, but the airline later refused to grant the discount.
British Columbia’s Civil Resolution Tribunal found Air Canada liable for negligent misrepresentation: failing to take reasonable care about misleading information the passenger reasonably relied on, causing financial loss.
In Mata v Avianca (2023), lawyers submitted court papers citing cases that ChatGPT had invented. A New York federal court penalised the lawyers and their firm under existing court rules.
The decision addressed failures to verify the cases and subsequent misleading statements to the court. The penalty was not imposed merely for using AI. Professional responsibility required checking the work before presenting it as reliable.
Recruitment also raises questions about suppliers. In a July 2024 ruling in Mobley v Workday, a federal court in California allowed certain claims of discriminatory applicant screening to proceed. Derek Mobley alleged that Workday’s tools disadvantaged applicants on the basis of race, age and disability.
The court accepted that Workday could be treated as an employer’s agent because it performed recruitment tasks on the employer’s behalf. However, the ruling did not establish that discrimination had occurred.
Foreign decisions offer useful comparisons, but Namibian courts must apply the law governing the dispute. Differences in legislation or facts may render a foreign court’s reasoning unsuitable.
How Namibian law could apply
In Namibia, a chatbot’s statement about a discount could form part of a contract if the requirements for an agreement are met.
Refusing to honour it could then constitute a breach of contract. Other harm may fall under delict, the law that allows compensation for certain civil wrongs without relying on a contractual promise.
A delictual claim requires more than proof of an error. The court must examine wrongfulness: whether the law recognises a duty not to cause that kind of harm in those circumstances.
It must also consider fault, such as intention or negligence, and whether the conduct caused harm for which compensation can be awarded. These are separate questions; carelessness does not automatically satisfy all requirements.
The Constitution also matters. Article 10 protects equality before the law and prohibits discrimination on grounds such as race or ethnic origin. Article 13 protects privacy, including communications.
Article 5 requires state institutions to respect and uphold these rights. It also binds individuals and companies to whom these rights apply.
In the loan example, a court would need to establish why applicants from one region were rejected while others with similar financial circumstances were approved.
It would then assess whether that treatment violated the right to equality under Article 10. A pattern of rejections could support a discrimination claim, but it would not, on its own, prove unlawful conduct by the institution.
Preparing judicial officers to assess the evidence
Applying these rules requires understanding the evidence. Preparation begins by identifying the role AI played. Did it draft a message, recommend approval of a loan, or automatically reject an application?
If someone reviewed its recommendation, what did that person check? Judicial officers need sufficient understanding to follow the evidence on the information used, the result produced, and the people involved.
The evidence must address the disputed decision. For a loan application, relevant records may include the applicant’s financial information, the bank’s lending rules, the tool’s recommendation and staff checks.
Other applications may show whether the bank treated comparable businesses differently. A supplier’s statement that its system is generally accurate does not explain why this applicant was rejected.
An expert should explain how the tool was tested, what the tests show, and what they cannot establish.
Did the testing include businesses from the regions and with the financial circumstances involved in the dispute?
A chatbot-generated explanation may sound convincing without reflecting the original decision. The court needs evidence linking the explanation to what actually happened.
Human involvement also warrants examination. A signature approving an AI recommendation does not show that anyone properly checked it.
Did the reviewer have the information, time and authority to question the result? Could they correct it, or were they expected to accept it? The answers help a court determine whether a person exercised judgement or merely passed on the tool’s recommendation.
Access to records may itself be disputed. A bank might resist providing other applicants’ information on confidentiality grounds.
The court would need to apply the rules governing access to documents and consider whether relevant evidence could be provided with identifying details removed where appropriate.
Both sides need a fair opportunity to challenge the evidence. Technical complexity should not be mistaken for reliability.
Courts must also be prepared for evidence created or altered using AI. If an audio recording is challenged, its source, storage history and possible alteration may need to be examined.
The judiciary of England and Wales provides a useful reference in its Artificial Intelligence (AI): Guidance for Judicial Office Holders, dated 31 October 2025. It warns that AI can fabricate legal cases and quotations and generate fake text, images and videos.
It also emphasises that judges must read the underlying documents themselves: AI cannot replace their direct examination of the evidence.
These lessons could inform Namibian judicial training on verifying legal references and assessing the authenticity of material presented in court, while applying Namibian law and court procedures.
Training could use a hypothetical loan dispute featuring sample applications, a supplier report and competing expert explanations.
Judicial officers could practise identifying missing records, questioning technical conclusions and determining what the evidence proves.
They should assess both sides fairly, without assuming the technology is reliable or the complaint justified and explain their decisions in language the parties understand.
How lawyers and compliance professionals can prepare
Preparation also extends to legal advice and compliance. Lawyers can help clients avoid disputes before adopting a tool.
For recruitment software, ask how the client will detect whether qualified applicants are being rejected for reasons unrelated to the job. For a chatbot, verify answers against actual prices and terms.
Supplier contracts should address access to records, assistance with complaints and responsibility for losses. A supplier’s promise to reimburse a business does not necessarily prevent a customer from suing that business.
When a dispute arises, preserve relevant instructions, responses, software versions and records of human checks before they are deleted.
Lawyers should verify their own AI-assisted work against judgments, legislation and supporting documents. Before uploading client information, establish who can access it, whether the provider may reuse it and whether sharing it is permitted by professional duties and the client’s instructions.
Compliance professionals help organisations follow the law and their internal rules. They should review actual AI use against the organisation’s AI usage policy: approved tools, permitted tasks, information staff may enter, and required human checks.
For example, if the policy prohibits uploading customer financial records to a public chatbot, staff need to understand the restriction, and compliance checks should confirm whether they follow it.
Review samples of completed work, record departures from the policy and refer them to the manager responsible for correction. Staff should know who can pause a tool or investigate complaints.
Repeated complaints should prompt investigation even if the tool previously passed its tests. Following an internal policy does not itself establish compliance with the law; the policy must also reflect the organisation’s legal obligations.
The role of boards and business owners
Where an organisation has a board, its directors’ duties include acting honestly in the company’s interests and exercising care and skill.
They should question management about AI uses that could harm customers, employees or the company, request records of testing and complaints, and follow up on unresolved issues.
AI failure does not automatically render directors personally liable. Any claim must establish what they did or failed to do and why the law makes them liable.
Organisations without boards also need supervision. Owners, partners or senior managers should assign responsibility for approving AI use, monitoring for problems and stopping a tool that is causing harm.
Preparing with international guidance
Preparation need not await dedicated Namibian AI legislation or comprehensive data protection legislation. UNESCO’s Recommendation on the Ethics of Artificial Intelligence sets out principles for fairness, privacy and human oversight.
ISO/IEC 42001:2023 sets out requirements for how organisations manage AI, including policies, responsibilities and risk controls.
These international standards and guidelines can help organisations use AI responsibly, but they do not automatically become binding Namibian law. Following them does not, by itself, determine whether an organisation is legally responsible for harm.
A dispute over a loan, a job application or professional advice could bring an AI-assisted decision before a Namibian court.
Judicial officers should be prepared to understand how the decision was made, assess the evidence, and apply Namibian law, drawing on relevant international standards and guidance where helpful.
Chisom Obiudo is an admitted legal practitioner of the High Court of Namibia specialising in corporate governance and AI governance. She is currently a Chief Legal Officer at the Namibian Law Reform and Development Commission.
