- Published on
How AI could ‘supercharge’ election risks across south-east Asia
- Authors
- Name
- DeepHorus
A disturbing new report has revealed how artificial intelligence was used to orchestrate a sophisticated political influence operation in Malaysia, raising urgent warnings about AI's potential to "supercharge" election interference across south-east Asia. Anthropic, the company behind the Claude AI model, disclosed that its technology was exploited to build a commercial election manipulation platform targeting voters during recent state elections. The operation deployed a network of approximately 1,000 fake X accounts, a fabricated news outlet called "Malaysia Pulse," and a series of forged documents designed to spread disinformation at scale. The revelation has alarmed cybersecurity experts and election observers across a region already grappling with persistent threats to democratic processes.
The operation, detailed in Anthropic's threat assessment report released in September, represented a dramatic evolution from earlier disinformation campaigns that once required vast teams of cyber operatives. Previously, such campaigns demanded an army of actors manually creating fake news websites, social media accounts, and forged dossiers. Now, AI tools can generate enormous volumes of persuasive, targeted content rapidly and at a fraction of the cost. Anthropic said the actors involved included governments, state-aligned propaganda institutions, state media, and private firms selling influence services commercially.
The Malaysia campaign pulled in census and electoral data to exploit sensitive political and social fractures, including race, religion, and royalty, across all 222 Malaysian parliamentary constituencies. The operation was designed to precisely target voter demographics in specific constituencies using deeply personalised messaging. Although Anthropic intervened before the campaign could achieve meaningful impact, its mere existence demonstrated the alarming scalability of AI-driven manipulation. The platform targeted audiences across six continents, suggesting a far broader ambition than Malaysia alone.
Technically, the operation leveraged Claude to automate the creation of fake social media profiles and convincing news websites. AI models can produce text that mimics authentic journalism, craft persuasive political narratives, and tailor messages to specific demographic groups with unprecedented speed. As Katie Harbath, former head of public policy for global elections at Meta and current head of consultancy Anchor Change, noted, "AI can supercharge it in terms of you can make a lot more content that's more personalised, quicker and cheaper." However, Harbath cautioned that distribution mechanisms through social media platforms remain essential for such campaigns to succeed.
The impact could be severe across a region home to predominantly young, hyper-connected populations vulnerable to viral disinformation. In the Philippines, Nobel laureate journalist Maria Ressa has described the country as a "petri dish" for testing disinformation techniques later replicated in Western democracies. Rossine Fallorina of Philippine thinktank Sigla explained that AI essentially mimics traditional "troll factories" by accelerating production, noting that one prompt can replace ten people creating narratives. Last year, OpenAI banned accounts using AI to generate political comments about Philippine officials, many in support of President Ferdinand Marcos Jnr.
Bridget Welsh of the University of Nottingham Asia Research Institute Malaysia warned that AI-driven content inherently feeds political polarisation and exploits existing social divisions. Dr Ben Loh, a media lecturer at Monash University Malaysia, said AI has been used to inflame tensions, including inciting hatred toward ethnic Rohingya refugees fleeing Myanmar's civil war. Indonesia already witnessed AI's influence during its 2024 election, when a viral AI-generated cartoon image helped soften President Prabowo Subianto's strongman image. These examples illustrate both the manipulative and persuasive potential of AI in electoral contexts.
Looking ahead, major elections loom across the region, with Malaysia's general election potentially called for late 2027, the Philippines and Cambodia voting in 2028, and Indonesia following in 2029. Governments and technology companies face mounting pressure to develop robust safeguards before these polls arrive. Regulatory frameworks across south-east Asia remain fragmented and often ill-equipped to address AI's rapidly evolving capabilities. The Anthropic revelation serves as a stark reminder that the window for preparation is narrowing fast, and that without coordinated international action, AI-driven election interference could fundamentally undermine democratic integrity throughout the region.
