Who’s Liable When the Algorithm Is Wrong? Ghana’s Corporate Governance Gap on AI

Ghana has no law that says who pays when an AI system makes a wrong decision. This article explains the...
Ghanaian boardroom table with a glowing AI symbol and a book on the Companies Act 2019.

This article is general guidance, not legal advice. It has not been reviewed by a qualified Ghanaian legal practitioner. It draws substantially on a single legal practice guide rather than on primary legislation, so it should not be used as the basis for a real legal decision. If your company faces a real claim, a regulatory inquiry, or a decision with serious legal risk, seek independent advice from a qualified Ghanaian lawyer before you act.

An AI driven credit scoring model at a Ghanaian financial institution denies a qualified small business owner a loan. Or an automated fraud detection system locks a legitimate customer out of her account for days. In either case, a simple question follows. Who is actually liable?

The honest answer, under Ghana’s current legal framework, is that it depends, mostly on laws that were never written with AI decision making in mind.

On 24 April 2026, President John Dramani Mahama launched Ghana’s National Artificial Intelligence Strategy, a ten year plan for 2025 to 2035, designed to speed up AI use across education, health, agriculture, finance and public services. At the same time, the Bank of Ghana’s Cyber and Information Security Directive, known as CISD 2026, already sets AI governance requirements inside banking. What has not kept pace is a clear framework for board level accountability when AI decisions go wrong outside the part of the economy that CISD 2026 directly governs.

What Is Actually Happening

An algorithm is a set of rules a computer follows to solve a problem or make a decision. In this context, AI means a computer system that learns patterns from data and uses them to make or support a decision. Credit scoring means judging whether a person or a business is likely to repay a loan. Fraud detection means spotting transactions that look suspicious and stopping them before money is lost.

In Ghana today, these systems are already at work. A bank may use AI to score a loan application or to flag a mobile money transaction. An insurer may use AI to assess a claim. A retailer may use AI to decide who gets credit on instalment.

When a human officer makes a wrong decision, it is usually clear who made it and what information they used. When an AI system makes a wrong decision, those questions become harder. Did the board approve the system? Did management test it? Did the vendor explain how it works? If no one can answer with records, the company is exposed, even if no one intended harm. See our guide on how mobile money fraud actually works in Ghana.

Ghanaian small business owner checking her phone in her market stall in Accra.
A wrong automated decision can mean a refused loan or a frozen account for a real business owner.

Why This Is Different From Older Business Risks

Boards are used to managing risk, such as credit risk, the risk that a borrower will not repay, and operational risk, the risk that a process or a system will fail. AI risk is different because it can combine several risks at once, at speed and at scale.

First, an AI error is rarely a single mistake. A flawed model can repeat the same error across thousands of decisions before anyone notices.

Second, the reason for the error is often hard to explain. A traditional rule is easy to trace. A manager can say that an application was refused because the income figure was below the threshold. A machine learning model, which is a type of AI that learns from examples rather than from fixed rules, may reach a decision through patterns that no single person wrote down.

Third, the harm can fall unevenly. If the training data reflects old bias, the system can repeat that bias in new decisions. Fairness is therefore a governance question, not only a technical one.

Who Is At Risk, and Who This Is For

This article is for directors, company secretaries, senior managers, founders, and compliance officers in Ghana, and for anyone who buys, builds, or relies on an AI system that affects customers.

The risk is highest for companies that use AI to make decisions about people: banks and other lenders, insurers, fintech companies, payment firms, retailers that offer credit, and employers that use AI to screen job applicants. It also includes companies that use a vendor system without fully understanding it. Buying a system does not remove responsibility for how it is used.

Small and medium businesses are not exempt. A small lender that uses a simple scoring tool still makes decisions that affect livelihoods. The scale is smaller, but the principle is the same. If your system decides, you need to be able to explain the decision.

Customers are at risk too. A wrong AI decision can deny credit, freeze an account, or delay a claim. For a market trader, a refused loan can mean lost stock and lost income. That is why this is not only a legal question. It is a question of trust.

Ghanaian company directors in a board meeting in Accra reviewing a report together.
AI oversight belongs at board level, with a named committee and a written record.

What Ghanaian Law Says

Ghana currently has no AI specific corporate governance law and no mandatory AI specific disclosure requirements for companies. What exists instead is a patchwork of general laws that apply to AI only indirectly.

The Companies Act 2019, Act 992, sets out the duties of directors. In plain words, a director must act in good faith, honestly and with the best interests of the company in mind, and must exercise care, skill, and diligence. Under those duties, directors may be personally liable for breaches of those duties, and for false or misleading disclosures relating to AI systems. That is a possibility to discuss with a lawyer, not a settled conclusion.

The Cybersecurity Act 2020, Act 1038, established the Cyber Security Authority and regulates cybersecurity activities in Ghana. Failures to protect critical systems, including AI systems, can trigger regulatory, civil, and criminal liability.

The Data Protection Act 2012, Act 843, governs personal data used to train or run AI systems. Personal data means information that can identify a living person, such as a name, a phone number, or a Ghana Card number. Data controllers must register with the Data Protection Commission before they process personal data. A person also has the right to ask for a decision made only by automatic means to be reconsidered.

Boards therefore face real exposure today, but through indirect routes. A director is not sued for AI governance failure as such. A director may face a claim for breach of the general duty of care involving an AI system, or for a data protection breach inside a machine learning pipeline. This coverage is real but unpredictable, and it leaves gaps, particularly outside finance.

What CISD 2026 Adds Inside Banking

The clearest exception is banking. In March 2026, the Bank of Ghana launched CISD 2026, which replaces the 2018 directive. It has six pillars, including AI and machine learning governance, board level accountability, and a proportionality framework that tailors requirements to the size and risk of each institution. It requires at least one board member with verified cyber risk expertise, and it applies to banks, microfinance institutions, savings and loans companies, fintechs, and payment firms regulated by the Bank of Ghana. For the detail, read our explainer on what CISD 2026 means for the wider digital economy.

Sector regulators can also act in their own areas, but they have no AI specific rulebook for companies outside their own sectors. That is why CISD 2026 matters, and why its limits matter too.

A Pattern Ghana Has Already Lived Through

Ghana’s own governance history offers a useful, uncomfortable precedent. The banking sector governance directives did not arrive as a proactive measure. They arrived after a wave of bank failures exposed how limited director accountability had been.

In August 2017 and August 2018, the Bank of Ghana revoked the licences of seven banks: UT Bank, Capital Bank, uniBank, The Royal Bank, Beige Bank, Sovereign Bank, and Construction Bank, citing insolvency and poor governance among the causes. It then issued its Corporate Governance Directive in December 2018, followed by a Corporate Governance Disclosure Directive in May 2022.

Ghana did not build strong director accountability into banking ahead of the crisis. It built it afterward, in response to one. AI governance is following the same trajectory. Just as pre crisis banking regulation assumed existing company law was sufficient oversight, today’s assumption is that existing corporate and data protection law is sufficient oversight for AI decisions. In banking, that assumption proved costly. For the banking side, see how Ghana is closing the AI adoption gap in financial services.

What To Actually Do: A Practical Checklist

The law may be unclear, but the governance work is not. A board does not need to wait for a new AI law to act.

  1. Add AI risk to the corporate risk register, a written list of the main risks a company faces. Treat it as its own entry, not as a small part of general IT risk.
  2. Assign clear board level ownership to an audit, risk, or dedicated technology governance committee. Record that decision in the board minutes.
  3. Document how each AI system works, in plain language: what data trains it, how outputs are reviewed, and what human oversight exists. Keep a log of tests, errors, and corrections.
  4. Test for fairness and accuracy before you scale, using real Ghanaian data. Check whether the system treats different groups differently without good reason. Record the results.
  5. Plan for mistakes and complaints. Write down how a customer can challenge an AI decision, who reviews it, and how quickly it will be answered.
  6. Use CISD 2026 as a design template, even outside banking. The CISD 2026 explainer sets out its six pillars.
  7. Check your vendor and your data protection position. Ask what data the system was trained on and how it is updated. Confirm registration with the Data Protection Commission.
Ghanaian compliance officer reviewing an AI system on a computer in a modern office in Accra.
Test the system, record the results, and keep a clear route for customers to challenge a decision.

Conclusion

Ghana has been here before, just with a different technology and a different sector. Banking sector governance arrived only after collapse forced the issue, at real cost to depositors, institutions, and public trust. AI governance is on a similar path, with rapid deployment running ahead of the accountability structures meant to govern it.

The National AI Strategy, launched on 24 April 2026, will speed up adoption. CISD 2026 already shows what governed AI looks like inside banking. The question is whether boards outside banking will learn from the banking clean up, or wait for an AI failure to force the same lesson. Boards that act now will write the next chapter on their own terms rather than in response to a crisis.

Sources Referenced

  1. Chambers and Partners, Corporate Governance 2026 Ghana Practice Guide, and Corporate Governance 2025 Ghana Practice Guide, Addison Bright Sloane. As cited in the original draft. The Ghana chapter authorship and the specific liability analysis were not independently re verified for this rewrite.
  2. Companies Act 2019, Act 992. Director duties, including good faith and care, skill and diligence, verified by browser search against DLA Piper, Global Guide to Directors Duties, Ghana, and Ghanaian legal summaries.
  3. Cybersecurity Act 2020, Act 1038. Text available from the Parliament of Ghana, Cybersecurity Act 2020, Act 1038. Assented 29 December 2020, establishes the Cyber Security Authority.
  4. Data Protection Act 2012, Act 843. Administered by the Data Protection Commission. Assented 10 May 2012, in force 16 October 2012.
  5. Bank of Ghana, Corporate Governance Directive 2018, December 2018, and Corporate Governance Disclosure Directive, May 2022. Verified against Bank of Ghana publications.
  6. Bank of Ghana, Cyber and Information Security Directive 2026, March 2026. Launch and six pillars verified against the Bank of Ghana Governor launch speech and Bank of Ghana reporting.
  7. Ministry of Communication, Digital Technology and Innovations, Ghana Launches National AI Strategy, 24 April 2026. Strategy period 2025 to 2035, eight pillars.
  8. Bank of Ghana, banking sector reforms and register of resolved institutions. Seven bank licences revoked in August 2017 and August 2018, verified against Bank of Ghana publications.
  • About
    Saviour Amevor

Leave A Reply

Your email address will not be published. Required fields are marked *