AI-powered tools are now embedded in everyday commercial life across Africa. ChatGPT subscriptions, generative image and video platforms, AI-assisted legal drafting tools, and automated customer service bots have moved from novelty to standard business input, and African markets are an increasingly visible part of that growth.

This expansion has outpaced a question indirect tax systems have yet to settle: how should VAT apply to AI-generated services? Unlike streaming content or e-books, where most African jurisdictions already have established digital service VAT frameworks, AI services are automated yet interactive, often produced with minimal human input, and span outputs from creative content to code and financial analysis. Whether an AI service qualifies as an electronically supplied service (ESS) and where it is taxable remains unsettled across most of the continent.

In this article, we review how African VAT frameworks currently treat AI-generated services, the compliance gaps that remain open, and what providers and consumers of these services should be doing now.

Classifying AI Services for VAT Purposes

Most African jurisdictions with digital service VAT rules define electronically supplied services broadly, drawing on OECD guidance or EU/UK frameworks: services delivered over a network, automated, and requiring minimal human intervention. On that basis, mainstream commercial AI tools, ChatGPT subscriptions, AI image generators, and writing assistants fit comfortably within existing ESS definitions. Delivery is automated, and outputs are generated algorithmically.

Complications arise at the edges. Hybrid services that pair AI output with human review may shift into professional services or licensed software categories instead of ESS. API access and enterprise licensing change the place-of-supply analysis depending on whether the recipient is VAT-registered. And AI outputs that constitute copyrightable work, written content, software code, music, raise a separate question of whether the transaction is a service supply or a transfer of intellectual property, which can carry different VAT treatment entirely.

These ambiguities are not unique to Africa, but they are sharper where digital VAT guidance is limited and administrative capacity to litigate disputes is constrained.

How African VAT Frameworks Apply Today

Kenya offers the clearest evidence that AI services are already inside the VAT net. From 1 May 2025, OpenAI began charging Kenyan ChatGPT users 16% VAT, citing Regulation 3 of the VAT Electronic, Internet and Digital Marketplace Supply (EIDMS) Regulations. The move sits uneasily alongside Kenya’s National Artificial Intelligence Strategy 2025–2030, which promotes AI adoption even as VAT raises the cost of access for students, developers, and startups.

South Africa’s position needs correcting against an earlier expectation: a planned VAT increase to 15.5%, and later 16%, was suspended by the Western Cape High Court in April 2025 and subsequently withdrawn. The standard rate has remained at 15% since. What changed is scope: April 2025 amendments to the Electronic Services Regulations excluded B2B supplies made by foreign non-residents exclusively to VAT-registered vendors, taking enterprise AI API sales to South African VAT-registered businesses outside mandatory non-resident registration.

Nigeria’s framework changed materially with the Nigeria Tax Act, 2025, which restructured the former Federal Inland Revenue Service into the Nigeria Revenue Service (NRS) and, from 1 January 2026, brought non-resident digital providers fully into VAT at an unchanged 7.5% rate, backed by new withholding and collection-agent mechanisms.

Botswana has enacted a VAT Amendment Bill that would bring non-resident “remote services,” including AI tools, into VAT at the 14% standard rate once a BWP 500,000 registration threshold is crossed, though the effective date is still to be confirmed by the Minister of Finance.

Cameroon and Côte d’Ivoire illustrate a related but distinct route. From 1 January 2026, Cameroon’s Finance Law subjects non-resident digital platforms meeting a Significant Economic Presence (SEP) test of FCFA 50 million in local revenue or 1,000 local users to a 3% levy on gross Cameroon-sourced revenue. This is a corporate tax instrument, not VAT, and providers should not conflate it with VAT compliance. 

Côte d’Ivoire already charges 18% VAT on non-resident digital services and, under its 2026 budget, layered on a 30% SEP corporate tax, capped at 10% of revenue, for providers meeting a similar threshold, creating dual VAT and corporate-tax exposure for AI platforms operating there.

Beyond these markets, the pattern holds: Ghana, Tanzania, Zambia, and others apply VAT or SEP-style rules broad enough to capture AI tools, even though most frameworks were drafted with streaming and e-commerce in mind.

Place of Supply: Where Is the AI Service Taxable?

African jurisdictions with digital VAT rules generally apply a destination-based test, taxing the supply where the consumer is located, consistent with OECD guidance on cross-border digital services. For most B2C AI subscriptions, this is straightforward: a Kenyan individual subscribing to a US-based AI tool receives a taxable supply in Kenya.

These scenarios complicate the analysis. Where an AI service is accessed through a marketplace, it is often unclear who is the deemed supplier: the developer, the marketplace, or a reseller. Where AI output is embedded into another product before reaching the end consumer, AI content folded into a media product, or AI-generated code built into a software application, the intermediate AI supply may have a different place of supply from the final product. 

And where a single global subscription serves users across several jurisdictions at once, identifying the actual place of consumption is a factual exercise. None of this has been addressed in published guidance in Kenya, South Africa, or Nigeria.

Compliance Priorities for AI Service Providers

Foreign AI service providers face a fragmented compliance landscape: registration thresholds, rates, and invoicing rules diverge sharply across the jurisdictions above, and simplified registration infrastructure remains less developed than in Europe or Asia-Pacific. Two priorities stand out for providers not yet compliant.

  • Run a market-by-market exposure analysis; thresholds, rates, and deemed-supplier rules differ materially between Kenya, South Africa, Nigeria, and the Francophone bloc, treating the first three as the highest near-term enforcement risk.
  • Build invoicing systems flexible enough for diverging thresholds and filing cycles, including the SEP corporate-tax layer in Cameroon and Côte d’Ivoire, which sits alongside, not instead of, VAT.

Emerging Gaps in Current Frameworks

Three issues remain unresolved across virtually every African VAT framework. The first is AI-generated intellectual property: where a user commissions a bespoke creative work, software module, or legal brief, the supply may be closer to a licence or IP transfer than a conventional ESS, and the two can carry different VAT treatment. 

The second is agentic AI systems that autonomously book travel, place orders, or manage workflows on a user’s behalf, which can generate multiple deemed supplies without a settled view on who carries the compliance obligation. 

The third is definitional: large language model outputs are generated afresh for each prompt, and whether this affects when a “supply” occurs has not been tested anywhere in Africa.

These gaps are not academic. As enterprise AI adoption deepens in financial services, media, legal, and healthcare, it will generate concrete compliance disputes.

Policy Direction

The forces point toward tighter integration of AI into African VAT systems. Revenue pressure is the most immediate: the African Development Bank and ATAF have repeatedly flagged the digital economy as an underperforming revenue base, with East African e-invoicing reforms cited as evidence of achievable gains. Enforcement capacity is rising too, as platforms’ e-invoicing and data-sharing arrangements with tax authorities expand visibility into AI consumption. 

Continued alignment with OECD destination-based principles, even though the OECD’s own Pillar One Amount A convention remains under negotiation. The African Union’s Digital Transformation Strategy and the AfCFTA Digital Trade Protocol point toward eventual harmonisation, though this remains aspirational for now.

Key Takeaways

AI-generated services are not beyond the reach of African VAT systems. Kenya’s enforcement of VAT on ChatGPT subscriptions from May 2025 made that concrete, and Nigeria’s and Cameroon’s 2026 reforms extend the trend. The harder questions about hybrid AI services, AI-generated IP, agentic transactions, and multi-jurisdictional subscriptions remain unresolved and should not be assumed to map cleanly onto existing digital service VAT rules.

The priority for foreign AI providers is a registration exposure review starting with Kenya, South Africa, and Nigeria, followed by monitoring of Botswana’s pending effective date and the Francophone SEP layer. For Africa-domiciled businesses consuming AI as a business input, the focus should be on confirming whether VAT is correctly charged, whether it is recoverable, and whether a reverse charge applies where a foreign supplier has not registered. The gap between AI adoption and tax policy response is narrowing faster than many businesses expect.