# Can AI influence the outcome of South Africa’s 2026 elections?

An episode of [Afternoon Drive with John Maytham](https://invisiblenetwork.ai/podcast/1212305258), 19 Oct 2026, 7:43.

Audio: https://p.podderapp.com/1951050295/traffic.omny.fm/d/clips/5dcefa8e-00a9-4595-8ce1-a4ab0080f142/e2113905-04ff-472c-9fff-a6cd00b63b20/35e40176-a763-4e0c-8f7b-b4e800ea9e9c/audio.mp3?utm_source=Podcast&in_playlist=cab9ce5d-42b6-4239-9177-a6cd00b6d1ff

John Maytham speaks to Kyle Findlay, co-founder of Murmur Intelligence, about the experiment, what AI models reveal about the political narratives they have absorbed, and how chatbots could shape voters’ perceptions ahead of South Africa’s 4 November local government elections. Afternoon Drive with John Maytham is the late afternoon show on CapeTalk. Presenter John Maytham is an actor and author-turned-talk radio veteran and seasoned journalist. His show serves a round-up of local and international news coupled with the latest in business, sport, traffic and weather. The host’s eclectic interests mean the program often surprises the audience with intriguing book reviews and inspiring interviews profiling artists. A daily highlight is Rapid Fire, just after 5:30 pm. CapeTalk fans call in to stump the presenter with their general knowledge questions. Another firm favourite is the humorous Thursday crossing with award-winning journalist Rebecca Davis, called “Plan B”. Thank you for listening to a podcast from Afternoon Drive with John Maytham Listen live on Primedia+ weekdays from 15:00 to 18:00 (SA Time) to Afternoon Drive with John Maytham broadcast on CapeTalk: https://buff.ly/NnFM3Nk For more from the show, go to https://buff.ly/BSFy4Cn or find all the catch-up podcasts here https://buff.ly/n8nWt4x Subscribe to the CapeTalk Daily and Weekly Newsletters: https://buff.ly/sbvVZD5 Follow us on social media: CapeTalk on Facebook: https://www.facebook.com/CapeTalk CapeTalk on TikTok: https://www.tiktok.com/@capetalk CapeTalk on Instagram: https://www.instagram.com/ CapeTalk on X: https://x.com/CapeTalk CapeTalk on YouTube: https://www.youtube.com/@CapeTalk567 See omnystudio.com/listener for privacy information.

## Transcript

[0:00] Speaker 1: In two days and two days and three weeks, two
[0:06] Speaker 1: days and two weeks, two days and two weeks. My
[0:08] Speaker 1: goodness me, it's two days and two weeks, 16 days time.
[0:10] Speaker 1: We vote in the local government elections. I don't know
[0:13] Speaker 1: if you've made up your mind. A, I'm going to vote. B,
[0:17] Speaker 1: I'm going to vote for party X and individual Y.
[0:21] Speaker 1: And if you haven't yet made a decision which party,
[0:25] Speaker 1: which individual to vote for, but you are determined to vote,
[0:28] Speaker 1: how are you going about deciding. Are you doing some research?
[0:31] Speaker 1: Are you using AI for research? Kyle Findlay, co-founder of
[0:36] Speaker 1: Murmur Intelligence, has with a regular co-writer, put something very
[0:41] Speaker 1: interesting on Daily Maverick at the moment.
[0:42] Speaker 2: Hello, Kyle. Hi, John.
[0:45] Speaker 1: It seems to me reasonable to say I really care
[0:50] Speaker 1: about the environment. Please, will you sum up the environmental
[0:54] Speaker 1: policies of the six major political parties in Cape Town
[0:58] Speaker 1: ahead of November 4. Would that be a reasonable use
[1:03] Speaker 1: of Claude?
[1:05] Speaker 2: Sounds like the perfect use case for what they're there for. Okay.
[1:10] Speaker 1: And is there evidence that people are using it in
[1:12] Speaker 1: that way?
[1:15] Speaker 2: Well, some of the evidence is things like 1 in
[1:18] Speaker 2: 10 people globally use the chatbot approach to understand their news.
[1:27] Speaker 2: And that number jumps to almost one in five for
[1:29] Speaker 2: young people. So not necessarily on who I should vote for,
[1:35] Speaker 2: but in terms of understanding my worldview, which feeds into
[1:39] Speaker 2: who I would vote for that aligns with my worldview,
[1:42] Speaker 2: we are increasingly using AI chatbots to inform our worldview
[1:46] Speaker 2: and our perceptions of reality.
[1:50] Speaker 1: And we're using something which.
[1:53] Speaker 1: I don't know how to describe AI. But we're using
[1:56] Speaker 1: something which is a what? Which is a very, very
[2:00] Speaker 1: sophisticated digital encyclopedia to tell us how to think about
[2:04] Speaker 1: the world in which we live.
[2:07] Speaker 2: Yeah, it's something that has distilled. It's gone out onto
[2:11] Speaker 2: the Internet and it's distilled vast, vast amounts of human knowledge.
[2:16] Speaker 2: And then it has... biases and tools built into how
[2:22] Speaker 2: it distills that down and what it chooses to reflect
[2:24] Speaker 2: back to us. And what happens is that those biases
[2:30] Speaker 2: can reflect in the answers we get. Because whenever you
[2:34] Speaker 2: get an answer from a chatbot, it's always an incredibly
[2:37] Speaker 2: succinct and very confident answer. In the past, if we
[2:42] Speaker 2: wanted to find something, we would have Googled it and
[2:44] Speaker 2: we would have synthesized the top 10 pages of results
[2:46] Speaker 2: or top 10 links of results to and come to
[2:49] Speaker 2: a conclusion ourselves. And at that time, Google still had
[2:52] Speaker 2: a control over what we were seeing, but there was
[2:56] Speaker 2: still work that we had to do. And now we
[2:59] Speaker 2: don't need to do that work anymore. We just get
[3:01] Speaker 2: given the answer. And what we just need to be
[3:04] Speaker 2: aware of is that that answer is someone's answer, not
[3:08] Speaker 2: necessarily the answer.
[3:10] Speaker 1: But sometimes we're so confused, an answer seems to sort
[3:14] Speaker 1: it out.
[3:15] Speaker 2: Oh, well, sure.
[3:16] Speaker 1: There's somebody who's got a view. And that's a relief
[3:19] Speaker 1: because I can't come to a view myself. I mean,
[3:24] Speaker 1: what's the potential? Let me rephrase that. What's the current
[3:28] Speaker 1: risk of having that versus the risk down the line
[3:33] Speaker 1: where we get more wrapped up in this stuff?
[3:36] Speaker 2: Well, there's good and bad things. So there's research that's
[3:40] Speaker 2: shown that when people interact with a chatbot around a
[3:45] Speaker 2: contentious issue, especially if they hold very strong, maybe extreme
[3:50] Speaker 2: views on that issue. Through interacting with the chatbot who
[3:54] Speaker 2: responds where they're at, doesn't respond in a judgmental way,
[3:58] Speaker 2: and through the iterative process of having a conversation with
[4:00] Speaker 2: someone who holds very strong views, there's research that's shown
[4:03] Speaker 2: that it actually takes people with those extreme views and
[4:07] Speaker 2: kind of moderates them and moves them into the center.
[4:10] Speaker 2: So if you think about social media, which has had
[4:12] Speaker 2: the opposite effect of polarizing us around extreme views, chatbots
[4:17] Speaker 2: might actually have a positive role in society for kind
[4:20] Speaker 2: of like, you know, lowering the temperature a bit. But
[4:24] Speaker 2: the other side of it is that there are only
[4:27] Speaker 2: like six or seven companies in the world that control
[4:31] Speaker 2: most of the chatbots. You're either OpenAI, Anthropic, Google, DeepSeek, Kimmy.
[4:37] Speaker 2: There's just a handful of companies that increasingly are the
[4:41] Speaker 2: gatekeepers or the arbiters of truth. And what other research
[4:45] Speaker 2: has shown is that if you look at the range
[4:49] Speaker 2: of responses that these models give over time, they don't
[4:52] Speaker 2: tend to be completely neutral. They actually embed a political worldview.
[4:58] Speaker 2: And funnily enough, most of them seem to be left center,
[5:01] Speaker 2: and Grok is actually the least left of center. So
[5:07] Speaker 2: you have, through having lots and lots of conversations and
[5:11] Speaker 2: millions and millions of conversations, you actually have... an ideology
[5:16] Speaker 2: being espoused to a population. Yeah.
[5:21] Speaker 1: Because when I was forming my political consciousness, if it
[5:25] Speaker 1: is at all formed, I'm still not sure that it is,
[5:28] Speaker 1: I had no choice but to read newspapers, listen to
[5:31] Speaker 1: the radio, and talk to other human beings about it.
[5:35] Speaker 1: I didn't have the option of asking Claude or Grok
[5:40] Speaker 1: how I should think about something. And I don't know
[5:44] Speaker 1: whether it's just because I'm old that I think the
[5:46] Speaker 1: old ways are better and that it's better to come
[5:48] Speaker 1: to a view through conversation and through consideration than it
[5:55] Speaker 1: is through reading a succinct summary of 19 billion, 472
[6:02] Speaker 1: million other conversations.
[6:05] Speaker 2: There's definitely something to be said about doing the hard
[6:07] Speaker 2: cognitive work yourself and reaching a conclusion through that, without
[6:12] Speaker 2: a doubt. What the potential is, is that we have
[6:16] Speaker 2: a society that's very brittle in terms of our knowledge,
[6:21] Speaker 2: because it's centralized in just a few places. And I'll
[6:24] Speaker 2: give you a really extreme example of that. In 2025,
[6:28] Speaker 2: when around the time that President Ramaphosa was visiting Donald
[6:31] Speaker 2: Trump in the White House, and he was ambushed with
[6:34] Speaker 2: the white genocide examples, for one day, Grok would insert
[6:41] Speaker 2: the topic of white genocide into every single response it
[6:44] Speaker 2: gave to any user. So if I was asking about
[6:48] Speaker 2: a pair of running shoes, it would say, you should
[6:51] Speaker 2: consider Nike Asics. And hey, by the way, have you
[6:54] Speaker 2: heard about the white genocide currently occurring in South Africa?
[6:58] Speaker 2: That was just a single addition to the system prompt
[7:04] Speaker 2: of a global AI model that introduced a topic into
[7:09] Speaker 2: everything that was discussed. And so in that way, you
[7:12] Speaker 2: can see how, um, how fragile these systems are to, uh,
[7:16] Speaker 2: who owns them. And that was what the article was
[7:19] Speaker 2: about basically was, um, they, they give good information, but
[7:23] Speaker 2: you just have to be aware that, um, you're not
[7:26] Speaker 2: going to know when it's telling you something false. We
[7:29] Speaker 2: will always, and it probably won't be false. It will
[7:30] Speaker 2: be biased, um, because of the way in which they
[7:33] Speaker 2: deliver it with such confidence and, um, Kyle
[7:38] Speaker 1: Finlay, thank you very much indeed, co-founder of Murmur Intelligence.

Shows are listed from Apple Podcasts' charts and directory, and Podcast Index where it is used. Every show and episode belongs to its publisher and plays from the publisher's own feed; this site hosts none of them.