The most revealing moment in my recent conversation with AI Diagnostics CEO Braden van Breda came almost as an aside.
His company has developed an AI-powered digital stethoscope that can help screen for tuberculosis by analysing lung sounds. It is precisely the sort of innovation South Africa should celebrate. But instead, the interview left me incensed at the ANC’s continued puritanical pursuit of narrow ideology over the broader public good and at the way ordinary South Africans will pay for this with their lives.
The AI Diagnostics stethoscope is locally built and is aimed at one of the country’s gravest public health failures, TB. It is affordable enough for frontline clinics and designed for use by nurses and community health workers rather than scarce specialists.
The device records lung sounds. The AI model developed by Van Breda and his small team analyses them. Within seconds a result can help determine whether a patient should be referred for confirmatory testing. In a country where late detection remains one of the central reasons TB continues to spread and kill at a devastating scale, this has the potential to save tens of thousands of lives.
Van Breda explains that the real breakthrough was not merely the stethoscope but the dataset behind it. “Unlike X-rays, stethoscope lung sounds [are] not typically recorded,” he tells me. “So we had to collect that data set ourselves.” AI Diagnostics set up a facility in Mitchells Plain, recruited participants from 33 surrounding public clinics and spent a year and a half building what Van Breda describes as “the largest tuberculosis-validated lung sound database in the world” — the foundation on which the model could be trained.
But there is a catch. For now, the model is hosted in the cloud. It requires connectivity. That practical detail should recast the entire debate about Starlink in South Africa.
The argument has too often been reduced to Elon Musk and his politics, his temperament, his inflammatory comments and his relationship with South Africa’s racial politics.
Much of that criticism is justified. Musk isn’t an easy figure to defend. But public policy is not a personality contest. The real question is whether South Africa is prepared to deny useful infrastructure to poor communities because the provider is politically unpalatable and because the state is unable to distinguish empowerment from a rigid ownership formula.
A serious state would ask what outcome best advances broad-based empowerment
South Africa’s TB burden remains appalling. AI Diagnostics’ own material cites World Health Organisation data showing that 249,000 people fell ill with the disease in South Africa in 2024, while an estimated 54,000 died. That’s double the annual murder rate. Worse still, a national prevalence survey found that 58% of those who tested positive for TB reported no symptoms. A screening system built around symptoms is therefore missing many of the very people it needs to find.
This is a disease where time matters. Late detection means late treatment. Late treatment means more transmission. More transmission means more illness, more deaths and more pressure on an already overburdened health system.
The economic and moral logic is obvious.
Instead, the country is stuck listening to the high priests of transformation defend their positions on grounds of ideological purity.
The Independent Communications Authority of South Africa (Icasa) has indicated that under the Electronic Communications Act individual licensees must meet the 30% ownership requirement by historically disadvantaged groups. Equity equivalent investment programmes may be recognised in broader empowerment policy, but full alignment in telecoms licensing would require legislative amendment. Communications minister Solly Malatsi says he intends to pursue such amendments, while parliament’s communications committee chair, Khusela Diko, has backed Icasa’s insistence on the existing legal position.
On the law, Icasa may well be right. But that is not the end of the matter. It is the beginning of the policy failure.
A serious state would ask what outcome best advances broad-based empowerment. Is it necessarily a 30% shareholding? Or could empowerment also mean connecting thousands of rural schools, supporting local service providers, training technicians, enabling small businesses and giving public clinics the broadband required to deploy diagnostic tools?
An equity equivalent programme is not some neocolonial trapdoor through which multinationals escape the country. Bismark Tyobeka of North-West University has pointed to the automotive sector, where global manufacturers were allowed to establish empowerment funds rather than carve up global ownership structures. In Starlink’s case the mooted alternative included connecting 5,000 rural schools and 2.4-million pupils. That is not “anti-BEE”.
There is, of course, a serious sovereignty question here, but it is not the breathless one being retailed by those who appear to believe Starlink will annex us from low-Earth orbit. Nor can the sovereignty argument be allowed to float free of evidence. Starlink is now available in well over 100 markets.
One recent mapping of Starlink’s own availability data puts active consumer service at about 166 countries and territories as of May, which means a very large number of governments, including many with serious national security establishments, have concluded that the risks are manageable rather than disqualifying. If the US, much of Europe, parts of Latin America and a growing number of African states have deemed the risk as manageable, South Africa should be capable of the same.
The Starlink debate is a test of whether South Africa’s GNU can govern in the interests of outcomes. It is, in the case of AI Diagnostics and the thousands who suffer the scourge of tuberculosis, a matter of life and death.
