Dear Directors, AI Will Not Take the Blame: Corporate Liability in The Age of Artificial Intelligence

IMAGINE THIS

The board of a Tanzanian company is considering acquiring another business for TZS 20 billion. Management uploads the target company’s financial information, projections and market data into an Artificial Intelligence system. Within minutes, AI produces an impressive analysis. It identifies projected growth, calculates potential synergies and recommends the acquisition. The directors approve the transaction. One year later, the acquisition becomes a commercial disaster. Material liabilities were overlooked, the financial assumptions were defective, and information relied upon by the AI was incomplete.

At the shareholders’ meeting, the directors are asked why they approved the transaction. Their answer is: “The AI recommended it.” It sounds modern. But does it provide a defence under Tanzanian company law? In my view, that is where the real legal discussion begins.

THE COMPANIES ACT APPOINTS DIRECTORS, NOT ALGORITHMS, TO RUN THE COMPANY

Artificial Intelligence can analyse information, identify patterns, calculate risk and make recommendations. But under company law, there is an important distinction between assistance in decision-making and legal responsibility for the decision made.

Section 185(1) of the Companies Act, Cap. 212 R.E. 2023 provides that when exercising powers or performing duties, a director must act honestly, in good faith and in what the director believes to be the best interests of the company. That duty belongs to the director—not management, an external consultant or an AI system.

Therefore, where AI recommends a transaction, the legal question is not simply whether the AI produced a convincing recommendation. The question is whether the directors themselves exercised their powers consistently with their statutory duties. AI may provide the information. The director must still provide the judgment.

THE AI GOT IT WRONG” IS NOT REALLY THE LEGAL QUESTION

Company law does not necessarily impose liability upon directors merely because a business decision turns out badly. Businesses involve risk. Investments fail, acquisitions underperform and forecasts prove inaccurate.

The more important question is: What did the directors do before approve the transaction?

Section 188 of the Companies Act, requires a director to exercise the care, skill and diligence that would be exercised in the same circumstances by a reasonable person having the knowledge and experience reasonably expected of a person in the director’s position, together with any special knowledge and experience that the particular director possesses. The better company-law question is therefore not, “Was the AI wrong?” It is: “Did the directors exercise the care, skill and diligence required by section 188 when deciding to rely upon the AI?”

AI DOES NOT REMOVE THE DIRECTOR’S DUTY OF CARE

Suppose AI recommends acquiring a company. A careful board may reasonably ask: What information was given to the AI? Who verified it? What assumptions did the system make? Were the financial statements independently reviewed? Was legal due diligence undertaken? Did the system consider litigation, regulatory exposure, tax liabilities, employment liabilities and contractual obligations? Directors do not have to become computer scientists. But section 188 requires them to exercise the degree of care, skill and diligence appropriate to their position. The more material the decision, the more dangerous it becomes for a director simply to say: “I relied on the AI.” Reliance is not necessarily negligence. Blind reliance may be.

A DIRECTOR CANNOT OUTSOURCE HIS STATUTORY DUTY TO AI

Boards routinely rely upon accountants, advocates, valuers, auditors, engineers and other professionals. There is nothing inherently improper about using external expertise. AI can similarly be a powerful decision-support tool. But there is a line between relying upon information and surrendering the decision itself.

Section 185 of the Company Act, places the obligation to act honestly, in good faith and in the company’s best interests upon the director. Section 188 separately imposes the obligation to exercise reasonable care, skill and diligence. Neither duty disappears merely because technology was used.

A director may outsource analysis. He cannot outsource his statutory responsibility. Once a board stop interrogating an AI recommendation and merely rubber-stamps it, the matter ceases to be only a technology issue. It potentially becomes an issue of directors’ duties under the Companies Act.

THE AI RECOMMENDATION MUST STILL BE USED FOR A PROPER CORPORATE PURPOSE

Section 187 of the Companies Act requires a director to exercise his powers for proper purposes. Imagine directors using an AI-generated valuation to justify issuing new shares, acquiring another company, disposing of corporate assets or entering into a related transaction. The fact that an AI system produced figures supporting the transaction cannot sanitise an otherwise improper exercise of corporate power. The legal question remains: For what purpose was the director’s power exercised? Technology cannot convert an improper purpose into a proper one.

AI CANNOT CHANGE WHOSE INTERESTS THE DIRECTOR MUST PROTECT

Section 185 of the Company Act, focuses the director on what he believes to be the best interests of the company. This becomes important where AI recommends a transaction that benefits a particular shareholder, parent company, director or related entity but may disadvantage the company itself.

The board cannot simply say: “The algorithm showed that this was the most profitable transaction.” The immediate legal question is: profitable for whom? The Companies Act recognizes limited circumstances in section 185 where directors of subsidiaries and joint-venture companies may have regard to interests beyond the immediate company, subject to the statutory conditions. Artificial Intelligence may analyze a corporate group economically as one enterprise. Company law does not necessarily do so.

WHAT IF THE AI PRODUCES FALSE INFORMATION?

Generative AI can produce information that appears convincing but is inaccurate or fabricated. Suppose an AI-generated due diligence report states that no material litigation exists against a target company. The board approves the acquisition, only to discover substantial litigation after completion. Again, the legal issue is not simply that AI was wrong. Under section 188, the question is whether reasonable directors exercising appropriate care, skill and diligence would have treated an AI-generated response as a substitute for proper legal due diligence. The greater the corporate consequence, the harder it may become to justify unverified reliance upon an automated answer.

THE BOARD MINUTES MAY BECOME THE DIRECTORS’ BEST FRIEND—OR THEIR WORST ENEMY

Section 151(1) of the Companies Act requires a company to cause minutes of proceedings at meetings of its directors to be entered in books kept for that purpose. Under section 151(2), properly signed minutes constitute evidence of the proceedings. That has important implications for AI-assisted decisions.

If a major acquisition later becomes the subject of litigation, the board minutes may show whether directors merely received an AI recommendation or whether they questioned management, examined its assumptions, considered legal and financial due diligence, discussed risks and exercised their own judgment. For significant transactions, board minutes should appropriately demonstrate not only the decision, but also the decision-making process. When everything goes well, that distinction may appear insignificant. When litigation begins, it may become crucial.

AI MAY ACTUALLY RAISE THE STANDARD EXPECTED OF DIRECTORS

Some directors may believe AI makes their work easier and therefore reduces their responsibility. The opposite may eventually prove true.

Section 188 of the Company Act, asks what care, skill and diligence would be exercised by a reasonable person in the director’s position. As sophisticated analytical tools become readily available, reasonable corporate decision-making itself may evolve. A director who once said, “I could not reasonably have known,” may increasingly face the question: “Why did you not check?” Technology therefore does not necessarily lower the standard of corporate governance. It may raise it.

THE BOARDROOM MUST REMAIN HUMAN

None of this means Tanzanian companies should avoid Artificial Intelligence. AI can improve financial analysis, fraud detection, compliance monitoring, risk assessment, due diligence and strategic planning. The law should not require directors to ignore useful technology. But directors must understand the difference between using AI and hiding behind AI.

Sections 185, 187 and 188 of the Company Act, remain concerned with the conduct of the director: good faith, proper purpose, and reasonable care, skill and diligence. Section 151 may then provide documentary evidence of what the board actually did. A Tanzanian director may legitimately say: “I considered the AI recommendation.” He should be much more cautious about saying: “I approved the transaction because the AI recommended it.” The first statement demonstrates technology being used as an input into human judgment. The second potentially suggests the surrender of judgment itself.

 

CONCLUSION: THE MACHINE CAN ADVISE BUT THE DIRECTOR MUST DECIDE

Artificial Intelligence will undoubtedly find its way into Tanzanian boardrooms. The Companies Act may not expressly mention algorithms, machine learning or generative AI. It does not need to do so before its existing principles become relevant.

Section 185 asks whether the director acted honestly, in good faith and in what he believed to be the company’s best interests. Section 187 asks whether corporate powers were exercised for proper purposes. Section 188 asks whether the director exercised the required care, skill and diligence. Section 151 ensures that what happened in the boardroom may be recorded and later proved.

AI changes the tools available to directors. It does not change who the director is. So, when a major acquisition collapses and shareholders ask who is responsible, pointing towards a computer will not necessarily answer the legal question. The algorithm did not accept appointment as a director. The algorithm did not assume statutory duties under the Companies Act. The algorithm did not sit at the board meeting. The directors did. The machine may make the recommendation. The responsibility remains in the boardroom.

Pongolela David, Advocate

Corporate & Commercial Lawyer