AI policy blunder highlights risks in SA’s tech strategy

You can also listen to this podcast on iono.fm here.

JEREMY MAGGS: South Africa’s first Draft National Artificial Intelligence (AI) Policy has been withdrawn, as I understand it, after it emerged that the document contained fake, apparently AI-generated references. But the real question is whether government has corrected a mistake or has created a bigger one. Was this a sensible act of accountability, or maybe a policy retreat at exactly the moment when AI investment, data centres, cloud infrastructure, DeepSeek, Microsoft, AWS (Amazon Web Services) and foreign technology power are moving faster than the state.

Well, Nathan-Ross Adams is an AI law researcher and founder of ITLawCo. Before we talk to him, I want to hear very quickly from the Communications and Digital Technologies Minister, Solly Malatsi.

SOLLY MALATSI: Following revelations that the Draft National Artificial Intelligence Policy published for public comment, contains various fictitious sources in its reference list. We have initiated internal questions which have now confirmed that this was the case. This failure is not a mere technical issue, but has compromised the integrity and credibility of the draft policy.

As such, I am withdrawing the draft national policy. South Africans deserve better. The department did not deliver on the standard that is acceptable for any institution entrusted with the role to lead South Africa’s digital policy environment. The most plausible explanation is that AI-generated citations were included without proper verification. This should not have happened.

In fact, this unacceptable lapse proves why vigilant human oversight over the use of artificial intelligence is critical. It’s a lesson we take with humility, and I want to reassure the country that we are treating this matter with the gravity that it deserves.

Read:
Government pulls draft AI policy for using fake research
Stafford Masie rips government’s draft AI policy

JEREMY MAGGS: So, Nathan-Ross Adams, a very warm welcome to you. Fake citations in a national policy. Not a small mistake. Why should the public, do you think, trust the rest of the document?

NATHAN-ROSS ADAMS: That’s a really great question. I think what it really comes down to, at the end of the day, is whether the policy itself was AI-generated and not verified, or whether it was merely the reference list. So the citations that were attached at the end of the policy, whether those were not verified. Based on the minister’s comments and what’s available publicly, it was only the reference list.

So yes, it may have created a trust concern, but I don’t think it warranted withdrawing the policy.

I think what should have happened instead is that it’s the investigation consequence management, and then perhaps just an updated reference list published in the [Government] Gazette with the comment period maybe extended for another 30 days.

JEREMY MAGGS: So withdrawal is the wrong call, you’re suggesting. Maybe letting government off too lightly, though? What’s your view?

NATHAN-ROSS ADAMS: Oh, definitely not. I’m not suggesting that there shouldn’t be consequences for whoever actually included the references and why there was no oversight for that. But what I am saying is that when the policy is withdrawn and another one needs to be drafted to replace that, there are other commercial concerns that are taking place right now.

For example, without a framework, South Africa’s counterparties, the big tech businesses that we’re doing business with such as Microsoft, Huawei, Google, AWS, we’re accepting contracts on their terms, and we need a policy to cover that. So it becomes a weighing all the pros and cons and then making a decision that’s in our country’s best interest.

Listen/read:
SA’s draft AI policy: Can it boost innovation?
Recent advancements in artificial intelligence models
AI in Action: Google sees opportunities in SA
Amazon, Prosus reach AI, cloud deal for double-digit cost saving
Microsoft’s massive R25.8bn SA investment

JEREMY MAGGS: But Nathan-Ross, it still raises the question, if the policy itself could not verify its own sources, it does say something about the state’s ability, surely, to regulate AI.

NATHAN-ROSS ADAMS: Excellent question. A lot of commentary that has emerged publicly has been about, well, if we can’t trust the actual reference list, can we then snowball on and not trust the integrity of the full document? It’s the right question and it’s the right approach.

However, the solution or remedy to that is not withdrawal, in my opinion. It is let’s figure out what actually caused this, let’s find out what was unverifiable and we disclosed it then publicly, we update the correct list, but also clarify what parts of the policy are legitimate. There are ways to do that by producing the correct sources that were referred to. I think the response was too heavy handed, in my opinion.

JEREMY MAGGS: Time in this debate is now really of the essence here, because you rightly warn of a governance vacuum, and there are real world decisions that are being made right now while that vacuum exists, surely?

NATHAN-ROSS ADAMS: Agreed. Connected with the policy that many people aren’t talking about is the fact that the National Treasury released the procurement regulations for government generally. So that’s effectively how to do business with government.

One of the items that was completely missing in there, and which was completely missing in the draft policy itself, was this idea of AI procurement and locking us in as a country with foreign vendors.

So let’s say we’re limited to technology from a specific country, and each of our departments subscribe to them. That’s a problem for us because it means that from a national security perspective, we’re at risk. It means that if prices inflate, we’re locked in with a particular vendor. So there are a lot of real consequences for each of us.

Read: Trial of the century to determine who controls AI’s future

JEREMY MAGGS: There’s huge potential as far as South Africa is concerned. You suggest that we have leverage because of minerals, because of energy potential. There’s huge growth in data centres. The problem, though, as I see it, is that we’re not maximising that leverage enough, surely?

NATHAN-ROSS ADAMS: I agree with you completely. I think when we think about AI – and these were part of our submissions on the draft AI policy – it’s pointing to the fact that AI isn’t just a tool that’s independent of an existing supply chain. It consists of what’s known as semiconductor or computer chips, which effectively make up AI.

We supply most of the platinum group minerals that will actually produce that.But we’re not thinking of it from a vertical level, we’re just focusing on the technology itself.

ADVERTISEMENT:

CONTINUE READING BELOW

Because of our minerals that we have, because of our data centre capability, which is the greatest on the continent, we have quite a bit of leverage that we’re not even considering, and that’s a problem for me because we need to do what’s in our country’s best interests and what will grow our economy at the end of the day.

JEREMY MAGGS: And that vertical thinking that you’re talking about is not just a Department of Communications issue, is it?

NATHAN-ROSS ADAMS: Agreed. There’s a geopolitical component to it which we can’t get into right now, but all of these considerations can be highlighted with greater interaction with the public, also including corporates and policy practitioners. There are several policy practitioners, like me, who have been focused on this for a very long time and, who weren’t consulted in drafting any of the correspondence that has taken place. So we have the great talent, we have the people. we just need to align everything.

JEREMY MAGGS: So just finally then, and it’s a hypothetical question, but as far as the minister is concerned, is it quickly republishing, you’ve already suggested that it would be difficult to start again, but could we build an interim emergency AI governance framework at this point just to tide things over?

NATHAN-ROSS ADAMS: So we’re not completely lost? We do have the broader AI policy framework which exists, which sets out these principles. But what we’re lacking, which was the gap this draft policy was trying to fill, is the operationalisation of that. So putting the principles into action, and I’m not sure how long that’s going to take.

From looking back, the last withdrawal that we had from a policy perspective was about the digital or remote working visa, and that was a quick process. The relevant minister acknowledged what had happened, made the changes, and it was effected within weeks. However, there’s been no indication from the Department of Communications and Digital Technologies on whether the same approach will be followed for this AI policy.

JEREMY MAGGS: Nathan-Ross Adams, thank you very much indeed, AI law researcher, founder of ITLawCo. I appreciate your time. Thank you.

Read/listen:
SA to offer points-based and Nomad visas in 30 days
Leon Schreiber on Home Affairs’ migration to machine learning and online services
SA takes a major step toward fixing work visa mess