Vodacom wants a human rights baseline for African AI. It should publish its own first.
A principle proposed for a whole continent carries weight only when its author shows how it applies to its own algorithms.
Sign up for my Substack daily AI newsletter https://johanosteyn.substack.com
Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/
Vodacom has done something most African companies leave to others. It has published a detailed view of how the continent should govern artificial intelligence, and it has anchored that view in human rights. The paper deserves a serious reading, and much of it deserves support. It also raises a question that every board publishing a policy position should expect: if this is the standard you propose for everyone, how do you meet it yourself?
CONTEXT AND BACKGROUND
The paper, "Governing Artificial Intelligence in Africa: A Principles-based Framework for Responsible Regulation", was developed with the law firm Bowmans and launched on 21 October. It proposes principles including human rights, transparency, accountability, human oversight, fairness, privacy, security and proportionality, a phased approach to regulation, and a role for regional economic communities in making rules interoperable. It urges governments not to adopt the EU AI Act wholesale, and proposes a lead coordinating body inside an existing ministry or regulator rather than a new AI agency.
The timing matters. South Africa withdrew its draft national AI policy in April after its reference list was found to contain sources that do not exist. An expert panel is now redrafting it, and TheStak notes that Heather Irvine, a Bowmans partner, sits on that panel and describes the paper as a lobbying document. I should declare that I have contributed to South Africa's national AI policy process as a working group member.
INSIGHT AND ANALYSIS
None of this makes the paper wrong. Industry has expertise and a continental footprint that many governments lack, and a human rights baseline is more than most companies volunteer. Lobbying is normal, and Vodacom was open about who wrote the paper.
The difficulty lies elsewhere. Ranking Digital Rights, which assesses the world's largest telecoms groups, found in June that nearly all of them now publish AI-related human rights commitments or principles, but most are high-level and non-binding. Only a few give concrete detail on how AI is overseen internally, or whether customers can opt out of their data being used to build algorithmic systems. Principles, in other words, are the easy part.
Vodacom's own human rights statement already commits it to regular human rights impact assessments, including for AI. Publishing what those assessments found, for the systems it uses with customers, would turn a policy proposal into a demonstration. Some detail will be commercially sensitive or could help fraudsters, so a summary of the risks found and the actions taken would be enough.
I have previously written about this in another setting, arguing that a function asking others for AI governance should first govern its own tools. The same logic applies to a company asking governments to regulate by principle.
The wider pattern explains why proof matters. Researchers at the AI Accountability Lab, analysing coverage of the EU AI Act and three global AI summits, recorded 249 instances of corporate influence over regulation, with the claim that regulation stifles innovation the most common narrative. That research concerns the large AI developers rather than African telecoms operators, and Vodacom is not accused of anything in it. Vodacom's call for proportionate, phased rules may be sound, but it will be read against that pattern unless the evidence comes with it.
IMPLICATIONS
For Vodacom, the remedy is simple and within its control: publish a summary of its AI human rights impact assessments before the South African redraft is finalised, and invite independent review of at least one high-risk system.
For other boards, the lesson travels. A policy position your company publishes becomes a commitment you will be measured against, by regulators, journalists and critics. Before approving one, a board should ask whether the company could pass the test it is proposing for others, and what evidence it could show tomorrow.
For the government panel, industry papers are welcome input. They should sit alongside independent research, civil society and the people affected, and contributors with commercial interests in the outcome should declare them in the record.
CLOSING TAKEAWAY
Vodacom has asked Africa to govern AI by principle, starting with human rights. That is a good place to start. The company now has an opportunity few authors of policy papers take: to show, with its own systems, what meeting that standard looks like. A continent deciding whose rules to trust will notice whether it does.
Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence's impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan's work at https://www.aiforbusiness.net
Comments