In an age where seeing is no longer believing, the Council for Scientific and Industrial Research (CSIR) is making an urgent and open-hearted plea: no single organisation can fight the tide of misinformation alone. As artificial intelligence reshapes how information is created, shared, and consumed, the line between authentic content and cleverly fabricated falsehoods has become disturbingly thin. Hyper-realistic AI-generated videos, voice clones, and manipulated images now travel through social media at lightning speed, leaving people unsure of what to trust. This is especially dangerous during elections, when the integrity of public information is the very foundation of democracy. Avi Moodley, Acting Research Group Leader for Natural Language Processing and Principal AI Specialist at the CSIR, spoke honestly about the scale of the problem during a virtual media roundtable. She did not sugarcoat the challenge. The issue is enormous, she said, and organisations trying to protect the information space simply cannot afford to work in isolation. Whether it is data, infrastructure, resources, or funding, every piece of the puzzle matters. Her message was not just a warning but an invitation: come join us, strengthen our network, and help us expand the fight. There is something deeply human about her concern. If researchers trained to analyse information feel exhausted by the sheer volume of content, she asked, imagine what ordinary South Africans are going through as they scroll through feeds overflowing with conflicting claims, not knowing what is real and what is machine-generated. That fatigue is real, and it is spreading.
Moodley went further, explaining that advances in AI have, to a significant degree, industrialised the spread of misinformation, disinformation, and malinformation—what experts now call information disorder. It is no longer just a few bad actors sharing rumours in corners of the internet. AI can be used to generate synthetic content at scale, to craft emotionally charged narratives designed to manipulate perceptions, and to amplify those narratives across platforms with terrifying efficiency. The result is an information ecosystem flooded with so much content that even the most careful reader can drown. This is not the same as previous elections, Moodley stressed. We have not seen this phenomenon before, and future elections will likely be even more chaotic if we do not act now. The old ways of spotting fake news—checking the source, looking for poor grammar, noticing odd images—no longer work when AI can produce flawless, persuasive content in seconds. That is why the CSIR insists that the response must be coordinated. Researchers, fact-checkers, journalists, tech companies, government communicators, and civil society all need to share insight, data, and resources. No one has the complete picture alone. Fighting misinformation is not a solo sport; it is a collective defence of a shared reality. The CSIR is not simply wringing its hands, though. It has been quietly building practical tools and systems to help make sense of the noise, and it is eager to put those tools into the hands of people who need them most.
One of the most important things the CSIR does is follow a well-worn, disciplined process during every election cycle. It begins by identifying the keywords and phrases that shape election-related conversations. From there, the team collects relevant information from social media platforms, using natural language processing to enrich the raw data and add layers of context. This enriched information is then run through the CSIR’s misinformation and unrest detection systems, which look for early warning signals that might indicate a coordinated disinformation campaign or a potential flashpoint of social tension. The same data feeds into network analysis, helping researchers understand how false narratives spread from person to person, community to community. Once the pieces are assembled, the findings are placed into dashboards that the CSIR showcases throughout the election period. These dashboards are not meant to sit on a shelf; they are designed to be shared widely with partners and through public platforms, including the Southern African Development Community, so that decision-makers and communicators can see emerging threats in near real time. Think of it as a digital early warning system for the information landscape. It does not simply label things as true or false in a blunt way. Instead, it tries to map the currents of conversation, highlighting where manipulation might be occurring and where vulnerable communities might be exposed to harmful content. This is painstaking, granular work, but it is vital for safeguarding the democratic process. The CSIR is not trying to police every post; it is trying to give society a fighting chance by seeing the patterns before they become crises.
One of the CSIR’s most promising innovations is the Informed Fake News Advisor Project. This technology is designed to evaluate individual pieces of information, compare them against credible sources, and trace their origin as far as possible. It does not claim to be a magic bullet that tells people what to believe. Instead, it provides indicators—helpful signals that flag potential misinformation so that human beings can make informed judgments. Moodley was careful to explain that this tool is not necessarily meant for ordinary South Africans to use in their daily scrolling. Rather, it is meant for fact-checkers and journalists on the front lines, the people who wade through enormous volumes of content every day to identify what is problematic and deserves closer investigation. These are the workers who need the most help, because they are trying to protect the public while also being overwhelmed by the same flood of information as everyone else. The CSIR is also working on an information triage system that takes the raw material from the Informed Fake News Advisor and adds other variables and inputs to help prioritise content for further assessment. In time-sensitive situations like elections, this triage capability could be a game-changer, allowing fact-checkers to focus their limited attention on the most dangerous and urgent claims rather than being spread too thin. Moodley acknowledged that this technology is still a work in progress. The upcoming elections, she said, will serve as a live testing ground. The goal is to learn, adapt, and strengthen the tools so that by the time the next national elections arrive, South Africa will be far better prepared for whatever emerges from the shadows of the internet.
Beyond text, the CSIR is also venturing into new and difficult territory: audio. Most efforts to combat misinformation in South Africa so far have focused on text-based content, largely because text is easier to search and analyse. But much of the information that people consume today comes in spoken form—voice notes on WhatsApp, clips on social media, radio broadcasts, and podcasts. Here, the CSIR faces a particular challenge rooted in the country’s rich linguistic diversity. South Africa has twelve official languages, plus countless dialects and informal ways of speaking. Converting audio into text for analysis is not only technically difficult but also risks losing meaning, nuance, and tone. A joke can become a threat. A sarcastic remark can become a factual claim. Sincerity can become mockery. To address this, the CSIR is developing technology that bypasses the need for conversion altogether, allowing researchers to search audio content directly and efficiently. This could be a real breakthrough in dealing with multimodal content, which is becoming increasingly prevalent as AI makes it easier to create realistic voice recordings and video clips. Moodley called it a potential game-changer, and it is easy to see why. If we can understand what is being said in audio without losing the texture of the human voice, we can better identify harmful narratives and respond more quickly. This work is still in its early stages, but it demonstrates the CSIR’s commitment to staying ahead of the curve. Misinformation does not respect conventional boundaries between text, image, and sound, and neither should the tools designed to combat it.
At its heart, the CSIR’s message is about shared responsibility and human connection. The media roundtable was hosted by the CSIR in partnership with the Government Communication and Information System (GCIS) and other key stakeholders, and the discussion revolved around fact-checking, public communication, and the role of science and technology in safeguarding the integrity of public information. But beneath the technical talk was a deeper concern for ordinary people. Every manipulated image, every misleading audio clip, every false story that goes viral chips away at trust—not just in the media or in politicians, but in one another. When we cannot agree on basic facts, democracy struggles. When people are overwhelmed by conflicting claims, they may simply tune out or retreat into echo chambers, making them even more vulnerable to manipulation. That is why the CSIR is not content to work quietly in a lab. It is reaching out, asking organisations, communities, and individuals to join the effort. The challenge is enormous, but so is the potential for positive change if we work together. The CSIR has the tools, the expertise, and the determination. What it needs now is partners who share that commitment. The fight against misinformation is not a battle that can be won once and then forgotten; it is an ongoing, evolving process that requires vigilance, creativity, and collaboration. By pooling resources and knowledge, South Africa can build an information environment where truth has a fighting chance. The call has been made. The door is open. The only question is who will step through it and help shape a more honest, more trustworthy future for everyone.