In the bustling corridors of Lagos’ financial districts and Abuja’s consulting hubs, Nigerian professional firms and companies are racing to harness artificial intelligence (AI), particularly generative AI (Gen AI), to streamline operations, enhance decision-making, and outpace competitors. From law firms drafting and reviewing contracts with AI assistance to accounting practices automating audits, the promise is tantalizing. Yet, as adoption surges, projected to contribute $15 billion to Nigeria’s GDP by 2030 according to PwC’s recent article, AI in Nigeria—the specter of ethical lapses looms large.
Two recent high-profile scandals from Australia’s Big Four firms offer stark warnings: Deloitte’s AI-riddled government report marred by fabrications and errors, and PwC’s catastrophic tax data leaks. These incidents underscore the perils of unchecked innovation, from hallucinated outputs to breached confidentiality. For Nigerian entities, they illuminate a path not just to build robust AI tools but to deploy them with integrity, fostering trust in an economy where regulatory scrutiny is intensifying under the Nigeria Data Protection Act (NDPA) and emerging AI guidelines from the National Information Technology Development Agency (NITDA).
The Deloitte Debacle: When AI Hallucinations Undermine Credibility
According to the Financial Review, Deloitte Australia faced a humiliating reckoning when a 237-page report commissioned by the Department of Finance for approximately $440,000 was exposed as riddled with AI-generated flaws. Intended to advise on government procurement reforms, the document contained fabricated footnotes, nonexistent citations, factual inaccuracies, and grammatical blunders: hallmarks of Gen AI “hallucinations,” where models like GPT variants invent plausible but false information. One cited academic paper didn’t exist; another reference linked to an unrelated blog post. The errors, uncovered by vigilant reviewers, prompted Deloitte to issue a partial refund and revise the report, eroding client confidence and sparking calls for AI governance reforms.
This fiasco isn’t isolated but a cautionary tale for Nigerian firms. In a context where resource constraints might tempt over-reliance on off-the-shelf Gen AI tools like ChatGPT for report generation, the risks amplify. Professional services in Nigeria from KPMG’s advisory arms to Deloitte & Touche, often produce high-stakes deliverables: tax analyses, due diligence reports, which demand precision. Unverified AI outputs can propagate misinformation, leading to regulatory fines or reputational damage. The lesson? Building ethical AI tools begins with rigorous validation layers. Nigerian companies should integrate human-AI hybrid workflows: AI for initial drafts, followed by multi-tiered fact-checking protocols. Tools like custom fine-tuned models on verified datasets, combined with plagiarism detectors (e.g. Turnitin integrations) and citation verifiers (such as Zotero APIs), can mitigate hallucinations. Moreover, adopting frameworks like the European Union Artificial Intelligence Act (EU AI Act), risk classifications in form of labeling high-risk applications like legal reporting, ensures accountability. By embedding ethical audits from the design phase, firms can transform AI from a liability into a verifiable asset.
PwC’s Tax Leaks: The Perils of Data Mismanagement in an AI Era
If Deloitte’s scandal highlights output integrity, PwC Australia’s 2023 tax leaks expose the foundational sin of data ethics,a bedrock for any AI system. In the case of the PwC tax leaks, senior partner Peter Collins shared confidential Treasury briefings on multinational tax avoidance measures with over 20 clients, enabling them to restructure operations pre-emptively. What began as internal knowledge-sharing snowballed into a betrayal of public trust, costing PwC $100 million in lost contracts, executive ousters, and a A$450,000 fine per involved partner. The affair, dubbed “Taxgate,” not only violated client confidentiality but also undermined the profession’s impartiality, prompting Australia’s tax authority to bar PwC from sensitive tenders for years.
For Nigerian professionals, this resonates deeply in an era of Gen AI, where models thrive on vast datasets often sourced from sensitive client information. Firms like Ernst & Young Nigeria or indigenous players in fintech auditing, handle troves of personal and corporate data governed by the NDPA’s stringent consent and anonymization rules. Leaking such data, intentionally or via AI training could invite NDPA penalties up to 2% of global turnover. PwC’s breach underscores that ethical AI deployment begins with uncompromising data governance. By adopting a privacy-by-design approach from the outset, organizations must map and classify all data flows, applying differential privacy techniques to safeguard individual records within training datasets. For deployment, organizations should implement role-based access controls (RBAC) and maintain immutable audit trails, using blockchain-inspired logging to ensure every AI query can be traced back to an authorized user or source.
Nigerian companies can draw from global standards like ISO 42001 for AI management systems, mandating ethical impact assessments before rollout. In practice, this means training staff on data minimization i.e feeding AI only what’s necessary and conducting regular penetration testing to thwart leakages. By prioritizing confidentiality, firms not only comply with laws but also build client loyalty in a market wary of data breaches, as seen in Nigeria’s 2023 Flutterwave cyber incidents.
Crafting Ethical AI Tools: A Blueprint for Nigerian Innovators
Armed with these lessons, Nigerian professional firms must proactively architect AI tools that embed ethics from inception, extending seamlessly to Gen AI capabilities. The first pillar is transparency, unlike Deloitte’s opaque AI use, tools should include “explainability” modules, using libraries like SHAP in Python to demystify decision paths. For Gen AI, this means watermarking outputs (e.g. via OpenAI’s classifier APIs) to flag synthetic content, preventing unwitting propagation of errors.
Bias reduction should also reflect Nigeria’s unique social and economic diversity. Most generative AI systems are trained on global data that lean heavily toward western cultures, which means they often overlook local realities, such as how Nigerians mix English and local languages (code-switching) in everyday communication and even in official documents. Firms should curate inclusive datasets, sourcing from NITDA’s open repositories or partnering with universities and apply debiasing algorithms from Fairlearn.
Building ethical AI also demands multidisciplinary collaboration. Ethicists, lawyers, and subject-matter experts should be involved from the outset as PwC’s experience shows how a siloed culture can enable breaches and ethical failures.
Regulatory alignment is non-negotiable. Building on the National Information Technology Development Agency’s (NITDA) foundational National AI Policy, the National AI Strategy 2024 reinforces the need for human oversight, accountability, and ethical governance in AI adoption. In response, Nigerian companies can demonstrate leadership by establishing internal AI ethics boards, embedding responsible-AI principles into operations, and conducting quarterly reviews of deployed tools to ensure ongoing compliance and transparency. To manage costs, firms can leverage open-source platforms such as Hugging Face to fine-tune large language models on anonymized internal data, preserving both privacy and intellectual property control. This proactive approach not only helps prevent ethical and reputational crises but also positions Nigerian firms as regional leaders in responsible AI, attracting foreign direct investment (FDI) from ethically driven investors.
Deploying Gen AI Ethically: Integrating into Workflows Without Compromise
Deploying generative AI raises the stakes for professional firms, demanding safeguards that scale with use. A phased approach, as seen in Deloitte Australia’s AI-related reporting missteps, underscores the need for pilot programs before full-scale deployment. Firms should start small, using Gen AI for low-risk, internal tasks like summarizing meeting notes or drafting internal emails and only expand to client-facing reports once the technology has been tested and benchmarked against human performance baselines through A/B testing.
Training and oversight are equally vital. The PwC Australia tax leak scandal revealed how cultural and oversight failures can compromise confidentiality and trust based on the Financial Times article, PwC Australia’s culture attacked in tax leak scandal report.
Nigerian firms can learn from this by mandating AI literacy and ethics modules for all staff. Training should include realistic simulations of ethical breach scenarios such as data leaks or misuse of confidential information to build awareness and accountability.
Clear usage policies are non-negotiable. Companies should prohibit unvetted generative AI tools in public-facing work and require transparent disclosure when AI contributes to deliverables. For example, firms might include a statement such as: “This analysis incorporates Generative AI reviewed under human supervision.”
To ensure accountability, firms can implement continuous auditing and observability tools, similar to LangChain’s tracing capabilities used in Gen AI model chains, which flag anomalies in real time and maintain audit trails for review.
In Nigeria’s increasingly hybrid and cloud-based work environments, firms must extend these safeguards to their vendors and technology partners. This includes vetting cloud providers such as AWS, Microsoft Azure, or Google Cloud for compliance with the Nigeria Data Protection Act (NDPA, 2023), especially around data residency, consent, and lawful processing. Contracts should include explicit data protection and AI governance clauses that ensure compliance with the Nigeria Data Protection Act (NDPA, 2023) and the principles laid out in the National AI Strategy 2024, such as data minimisation, access control, accountability, and transparency
Across Africa, some organizations are already showing how this can work in practice. For instance, South African law firm Bowman Gilfillan has integrated AI-assisted workflows with ethical checklists and human review protocols, reportedly reducing documentation errors by nearly 40%. Nigerian firms can adopt similar frameworks, embedding ethics into everyday operations rather than treating it as a formality.
Finally, ethical deployment must be measured and rewarded. Firms can link key performance indicators (KPIs) to compliance, accuracy, and transparency metrics such as audit pass rates or verified citation accuracy. Recognizing and rewarding teams that uphold ethical standards helps build a culture of integrity rather than one driven purely by revenue targets.
In the long run, ethical AI deployment is not a checkbox, it’s a competitive advantage. It enhances efficiency, prevents reputational and legal risks, and reinforces the trust that defines Nigeria’s professional services sector. By learning from global missteps like Deloitte’s AI errors and PwC’s governance failures, Nigerian firms can establish themselves as leaders in AI ethics and responsible innovation across Africa, attracting foreign direct investment (FDI) from value-driven global partners who prioritize transparency and trust.
Conclusion: From Cautionary Tales to Ethical Triumph
The Deloitte and PwC scandals, though oceans away, mirror pitfalls Nigerian firms can sidestep through deliberate ethical AI stewardship. By building transparent, bias-aware tools and deploying them with rigorous oversight, companies can unlock Gen AI’s potential responsibly. In doing so, they not only safeguard reputations but also contribute to a trustworthy digital economy, one where innovation serves society, not subverts it. The time to act is now; the cost of inaction, as Australia learned, is unmeasurably high.
Written by Adeola Osifeko LLB, BL,LLM, ACIS, ABR, Principal, AEO Law Practice
The Legal Trove 