AI could spread election misinformation on an ‘industrial scale’

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

JEREMY MAGGS: As South Africa heads towards the local government election, it’s just around the corner, AI-generated voices, fake videos, spoofed messages could make political misinformation a whole lot harder to detect.

COMRiC, the Communications Risk Information Centre, it brings telecoms operators and other stakeholders together to identify things like crime and fraud, and share intelligence on emerging threats, is stepping into this debate.

Essentially, the question is, could this technology already being used by fraudsters now be turned on voters?

The chief executive officer of COMRiC is Advocate Thokozani Mvelase and joins me on the programme now. Advocate, welcome to you. How big a worry, how big a concern is all of this ahead of the election?

THOKOZANI MVELASE: I think the background always starts that we know the adoption of the usage of AI, that it far surpasses the governance or rather the ethical usage of it.

While it has good intentions, but what we’ve seen in terms of the issues of voice cloning, the spoofing, how AI can be used to commit various other crimes, even fraud.

But when we now have to talk about, for instance, the elections, you could also see how easy it can be used to manufacture and proliferate misinformation and disinformation on an industrial scale and manipulate images, the voices and videos in ways that it can make it impossible for anyone to distinguish the actual voices or the actual images or the message.

This could be used by individuals, but also, it could be used by political parties that might want to discredit other political parties. So that’s the danger we have when it comes to the use of AI.

ADVERTISEMENT

CONTINUE READING BELOW

JEREMY MAGGS: And that phrase that you use, industrial scale, is very worrying. I’m assuming that something that is manufactured could spread very quickly before someone denies or disapproves it.

THOKOZANI MVELASE: I fully agree. If we take the fact that we live in the world of the digital age where most communication happens via devices, whether SMS or WhatsApp most of the time, and you see how a number of us belong to various WhatsApp groups, it takes just one WhatsApp group to send a message and no one has verified it and it gets forwarded multiple times and that information has spread significantly without being verified, it’s spread to various individuals.

And that’s the danger that we have.

JEREMY MAGGS: You’re putting out a very timely warning at this point. Is there any evidence that these techniques are already being tested in South African political messaging, or are you simply getting ahead of a threat?

THOKOZANI MVELASE: I think it’s important to get ahead of the threat. One of the other things as well is that so often it’s very often difficult to identify what is false and what is true with the use of AI in this day and age.

It might be possible already that there are messages that are coming out that have not been verified, that might just be used as intention.

But where we come from, closer to the elections, especially this month of October, because the election is just at the beginning of November, we might see an uptick of various messages.

Some of these messages, we want to warn the communities or the individuals that they might even spark an unnecessary community protest, because local elections are really about service delivery.

Any message that could be used as misinformation through AI, discrediting one party against the other, can spark some unnecessary violence that we don’t want to see as a country.

JEREMY MAGGS: Advocate, COMRiC tells voters to “stop, verify and don’t amplify”. But in a fast-moving election campaign, I’ve got to ask you whether that is feasible? It’s difficult, surely, for an ordinary voter to verify a voice note or even a video that sounds exactly like someone they know.

ADVERTISEMENT:

CONTINUE READING BELOW

THOKOZANI MVELASE: Yes, I fully agree with that. Hence, I would caution against forwarding messages that you’re not the author of. Let’s start it there, because there’s a tendency of forwarding information or messages, and you’re not the author of it, and you don’t know whether it’s valid or not.

Also, my inclination is that the IEC (Independent Electoral Commission) that is responsible for making sure that we have free and fair elections, needs to build capability to make sure that they are able to detect some of these messages and be able to send counter messages to say, please ignore or please [note] this is not a true message.

If the IEC doesn’t have that capability, surely then we might have challenges because they’re supposed to monitor this, because later this might get contested in terms of the information.

So my urge as well is that if you’re not sure, even report it to the IEC to say, we’ve received this message, is it true or not, before you even forward it.

JEREMY MAGGS: This is probably a little out of your lane, but you talk about concern over capability. Do you think the IEC, political parties and for that matter, even the telecom companies, are genuinely prepared for this? Or do you think we might be going into this election behind the technology?

THOKOZANI MVELASE: I can – and it might not be with confidence – say there is a level of preparation. What we’ve known and what we’ve seen around the use of AI, no one is an expert. So there is a base preparation by political parties, by the IEC that I believe is in place.

However, at the same time, as we know, there is no expert to this. You can only prepare so far, especially in the world where cybercrime has just been seeing an increase, even with whatever security you might have in your platforms.

So that’s all we are saying, we need to be aware of it and making sure that there is a level of detection that happens in advance or even post facto, so that it gets reported and can be verified.

ADVERTISEMENT:

CONTINUE READING BELOW

Where we have information that disproves it, it needs to be communicated immediately so that it doesn’t spread further.

JEREMY MAGGS: Is there an argument that maybe political parties should be required to authenticate official digital messages in some way or another? Again, easy to say, but maybe difficult to do.

THOKOZANI MVELASE: There should be actually. I think also with part of the pledge that the political parties signed, towards the election, it should also talk about a responsible campaigning process.

Because anybody who could be found in identifying that they’ve used misinformation through the use of AI, should be held accountable, actually.

That’s where probably we will talk about ethical usage of the AI platforms to spread information. Because now we’re going to be getting a lot of these calls which are bot-generated, and sometimes you never know whether it’s true or not.

JEREMY MAGGS: Just final one, Advocate, do you think there is a degree of responsibility that should be carried by the platforms themselves – WhatsApp, Meta, TikTok – when manipulated content, as we’ve been discussing, does go viral?

THOKOZANI MVELASE: I’ve always advocated for that, that there needs to be a level of responsibility and accountability by these platforms because it cannot go on uncontrollably. It goes to the verification of the information loaded into their platform and the people who are using their platform.

So that should there be issues of disinformation that might be critical, that it needs to be investigated, we should be going to them as a first call to verify.

JEREMY MAGGS: I’m going to leave it there with my thanks, Advocate Thokozani Mvelase, you are the chief executive officer of the Communications Risk Information Centre. Appreciate your time. Thank you.

#spread #election #misinformation #industrial #scale