Years ago, spreading disinformation on a national scale was a grueling, clandestine operation. You needed an army of cyber troops, a constellation of fake social media accounts, phony news websites, forged dossiers, and endless hours of coordination. Today, all it takes is an AI prompt. That unsettling reality came into sharp focus when Anthropic, the company behind Claude, revealed that its own AI had been used in a political influence operation in Malaysia. According to the company’s threat assessment, bad actors used Claude to assemble a commercial election manipulation platform: a network of roughly 1,000 X accounts, a fake news outlet called “Malaysia Pulse,” and a stack of fabricated documents. The operation was designed to target voters in recent state elections, pulling in census and electoral data to exploit the country’s most sensitive fault lines—race, religion, and royalty—across all 222 Malaysian parliamentary constituencies. In the end, the operation was ineffective once Anthropic stepped in, dismantling the network and strengthening its safeguards. But for those who study disinformation, it was a glimpse of a frightening future. AI can supercharge fake news in ways that were unimaginable just a few years ago, and it can do so from a laptop, with no army, no government budget, and no visible trace.
To understand why this matters, consider how disinformation used to be manufactured. In the past, a government or political operative had to hire people, often across multiple time zones, to craft messages, maintain personas, and push narratives. It was expensive, labor-intensive, and prone to mistakes. Katie Harbath, who once led public policy for global elections at Meta and now runs the tech consultancy Anchor Change, says AI has fundamentally changed the economics of manipulation. “AI can supercharge it in terms of you can make a lot more content that’s more personalised, quicker and cheaper,” she explains. “But you still need a distribution mechanism, and at the moment, that’s still your more social media type platforms.” In other words, AI doesn’t replace the entire machinery of disinformation—it makes the heavy lifting effortless. You can generate thousands of tailored posts in minutes, each designed to appeal to a different audience’s fears and prejudices. But you still need somewhere to spread it, and the big social platforms remain the battleground. This is especially concerning in South-East Asia, a region with a huge, young, hyper-connected population and a long history of being on the front lines of the fake news wars. Nobel laureate Maria Ressa, the Filipino journalist, has called her own country a “petri dish” for the weaponisation of disinformation. The techniques tested and refined there, she warns, later get exported to the rest of the world. So when we talk about AI and disinformation, this is not an abstract worry for Silicon Valley—it is a live, ongoing crisis for millions of people who get their news from social media.
The Philippines is not just a cautionary tale; it is a laboratory. Over the past decade, social media in the country has been flooded with troll armies, fabricated scandals, and coordinated harassment campaigns. Now, AI is making those efforts even easier. Rossine Fallorina, a researcher at the Philippine think tank Sigla, sees a direct line from old-school troll factories to the new AI-driven approach. “Basically AI is just mimicking what we would have, I guess, termed a few years ago, a troll factory,” she says. “Mostly because it speeds up the production process. So you don’t need 10 people to create a narrative or a story. You can just prompt it and then have it propagated.” Last year, OpenAI banned accounts that were using AI to generate comments about Philippine politics and public officials. Many of the fictitious comments were supportive of President Ferdinand Marcos Jnr., a sign that even a relatively small number of people can now manufacture the illusion of widespread online support. But the real danger is not just quantity—it’s the way AI can feed off existing divisions. “It feeds political polarisation and it feeds on divisions,” says Bridget Welsh, from the University of Nottingham Asia Research Institute Malaysia. “That’s really what’s concerning about the fragmented nature of what’s happening in Malaysia.” Across South-East Asia, where social media is often the primary source of news for young people, AI-generated content can make it harder than ever to know what is real, what is manipulated, and what is simply noise. The technology is not neutral; it is being used to exploit the deepest wounds in society, and it is working.
The stakes are high, and the clock is ticking. Malaysia could hold its next general election as late as 2027; the Philippines and Cambodia are scheduled to go to the polls in 2028, and Indonesia will follow in 2029. In the meantime, AI has already shown what it can do. During Indonesia’s 2024 election, an AI-generated cartoon image of Prabowo Subianto—a former general with a famously tough image—helped soften his public persona and became wildly popular online. It was a relatively benign use of the technology, but it demonstrated how easily synthetic images can shape political perceptions. Elsewhere, the effects have been darker. Dr Ben Loh, a media lecturer at Monash University Malaysia, points to AI-generated content that stokes hatred against Rohingya refugees who have fled to Malaysia to escape the civil war in Myanmar. The content doesn’t necessarily invent the hatred; it amplifies what is already there. “It’s come to the point where people already have this deep-seated hate in their minds and when they see the content like this, even though it’s AI-generated to them, it’s merely reinforcing what they believe is already true,” Loh says. This is perhaps the most insidious aspect of AI disinformation: it doesn’t need to convince anyone of something new. It only needs to reinforce what people already suspect, turning paranoia into certainty and prejudice into justification. And because the content is endlessly customizable, it can be tailored to different communities, languages, and grievances, making it even more dangerous than a one-size-fits-all propaganda campaign.
The Anthropic report that exposed the Malaysia operation offers a detailed look at how these schemes are built. The fake news outlet, “Malaysia Pulse,” scraped legitimate Malaysian reporting and had Claude rewrite the articles several times before republishing them under fabricated bylines. This gave the site the veneer of authenticity while stripping away the original source. The operation also republished articles from Russian and Chinese state-aligned outlets, including TV BRICS, Xinhua, Sputnik/RIA, and CGTN, but removed all attribution, making state propaganda appear to be independent journalism. Anthropic’s threat assessment says the actors behind such influence operations included “governments, state-aligned propaganda institutions and state media, as well as private firms selling influence.” The company said it “disrupted the activity, used what we learned to strengthen our safeguards, and shared intelligence with authorities and industry partners, where appropriate.” But the Malaysia case is likely just one of many. Because AI tools are cheap and widely available, any individual or small group can now engage in what used to be the domain of sophisticated intelligence agencies. They can generate fake news, manufacture social media personas, and even coordinate campaigns across multiple platforms without leaving their bedrooms. The only bottleneck is distribution, and even that is becoming easier as AI begins to infiltrate search engines, chatbots, and other services that people consult every day.
Looking ahead, Harbath warns of a new frontier: using AI to “flood the zone” in an attempt to manipulate the large language models (LLMs) that power AI assistants and search tools. If a flood of AI-generated articles, forum posts, and social media messages appears on the internet, it can pollute the data that future AI models learn from. “If there’s any sort of news drought, or if you have a place where there’s mostly government-sponsored media,” she says, “it wouldn’t take very much money to kind of flood the zone.” This could create a feedback loop: AI generates misinformation, other AI systems absorb it, and then serve it back as “truth” to millions of people. The result could be a world where the line between real and fake disappears entirely. But the story is not without hope. Anthropic caught this operation, and platforms are becoming more aware of the threat. Researchers, journalists, and civil society groups are paying closer attention. The challenge is to stay ahead of the technology, to build systems that can detect and dismantle these networks before they cause real harm, and to help ordinary people become more critical consumers of what they see online. In the end, the human element is still what matters most. AI can make disinformation faster and cheaper, but it still relies on our willingness to believe, share, and act on what we see. That is something no algorithm can fix—and something only we can guard against.