AI Could Intensify Election Risks in Southeast Asia

Published: 01 October 2026. The English Chronicle Desk. The English Chronicle Online

Artificial intelligence is creating a new dimension of risk for elections across south-east Asia, where young and highly connected populations rely heavily on social media for news, political discussion and public debate. Experts and researchers are increasingly concerned that the technology could make existing disinformation campaigns faster, cheaper and more sophisticated, allowing political influence operations to produce and distribute large volumes of convincing content with relatively limited resources.

The concern has intensified following the disclosure of an alleged political influence operation in Malaysia in which Anthropic’s Claude artificial intelligence system was used to support a campaign designed to manipulate political conversations. According to Anthropic’s latest threat assessment, the operation involved a network of about 1,000 accounts on X, a fabricated news organisation called Malaysia Pulse and a collection of false documents.

The operation appeared to be designed to reach voters during recent state elections by using census and electoral information to identify political and social divisions. Sensitive issues involving race, religion and Malaysia’s royal institutions were reportedly incorporated into the campaign across the country’s 222 parliamentary constituencies.

Anthropic intervened and disrupted the operation before it could achieve the intended scale of influence. Nevertheless, the case has raised questions about how similar technology could be used in future elections, particularly in countries where political debate is already heavily concentrated on social media.

The Malaysian operation illustrates an important change in the nature of political disinformation. In previous years, sophisticated influence campaigns could require large groups of cyber operators to establish fake websites, create social media identities, prepare fabricated documents and produce political narratives. Generative AI can now automate or accelerate many of those activities, reducing the time and resources required to produce content on a large scale.

Katie Harbath, who previously worked on global election policy at Meta and now leads the technology consultancy Anchor Change, has said AI can make disinformation more personalised, faster and cheaper to produce. However, the technology alone does not guarantee that misleading material will reach voters. Distribution networks, particularly social media platforms, remain an important part of the process.

That distinction is significant because political influence depends on more than simply producing large quantities of content. Material must reach an audience, attract attention and, in some cases, reinforce beliefs that people already hold. Social media algorithms, personal networks and existing political divisions can determine whether a particular piece of content remains largely unnoticed or becomes widely circulated.

South-east Asia already has extensive experience with online disinformation. The Philippines, in particular, has been closely studied by researchers examining the development of organised political influence campaigns. Filipino journalist and Nobel laureate Maria Ressa has previously described the country as an important testing ground for disinformation techniques, some of which have subsequently appeared in political environments elsewhere.

AI could make such techniques significantly easier to reproduce. Rossine Fallorina of the Philippine thinktank Sigla has compared generative AI with an automated version of the troll factories that previously relied on large groups of people to create and distribute political narratives. Instead of requiring numerous individuals to repeatedly write variations of the same message, AI can generate different versions rapidly and at relatively low cost.

The Philippines has already seen examples of AI-assisted political activity. OpenAI previously reported banning accounts that used its systems to generate comments concerning Philippine politics and public officials on social media. Many of the fabricated comments were reportedly supportive of President Ferdinand Marcos Jr.

Such incidents have added to wider concerns about how AI-generated material could affect political polarisation in countries where social and political divisions are already pronounced.

Malaysia is one of the countries where those concerns have particular relevance. Political debate can involve sensitive questions concerning ethnicity, religion and the monarchy, making certain subjects particularly vulnerable to manipulation. Bridget Welsh of the University of Nottingham Asia Research Institute Malaysia has warned about the interaction between AI-generated content and existing political divisions.

The concern is not necessarily that AI will create completely new political beliefs. In many cases, its influence may come from reinforcing attitudes that already exist. When people encounter content that appears to confirm what they already believe, they may be more inclined to accept or share it, even if the material has been generated artificially.

The potential use of AI in spreading hostility towards vulnerable communities has also attracted attention. Ben Loh, a media lecturer at Monash University Malaysia, has discussed the use of AI-generated material in relation to anti-Rohingya sentiment. Rohingya refugees have fled Myanmar amid prolonged conflict and persecution, and researchers have warned that online content can contribute to existing hostility towards the community.

The problem becomes particularly complicated when fabricated material is mixed with genuine information. Anthropic said the operators behind the Malaysian campaign used legitimate Malaysian news reports and instructed Claude to rewrite them before publishing the resulting material under fabricated bylines. The operation also reportedly reproduced articles originating from Russian and Chinese state-aligned media organisations after removing their original attribution.

Such practices can make it harder for ordinary readers to determine where a story originated and whether the apparent author or publication is genuine. A misleading campaign does not necessarily need to invent every fact. It can take legitimate information, alter its framing and present it through a false identity to create a different impression among audiences.

The ability to personalise political content is another area of concern. AI systems can generate different messages for different audiences, potentially allowing operators to focus on specific communities or demographic groups. If combined with electoral and population data, such technology could make political messaging considerably more targeted than conventional mass communication.

The region has already experienced both positive and controversial political uses of AI-generated media. During Indonesia’s 2024 presidential election, an AI-generated cartoon-style image of Prabowo Subianto became extremely popular online and contributed to a campaign effort that presented a softer public image of the politician. The example demonstrated how synthetic imagery could become part of mainstream political branding without necessarily being used as deliberately fabricated news.

The coming electoral calendar means these questions are unlikely to disappear. Malaysia could hold its next general election in late 2027, while the Philippines and Cambodia are scheduled to hold elections in 2028. Indonesia is expected to hold its next national election the following year.

By the time those elections take place, generative AI systems are likely to be more widely available and capable than they are today. Political campaigns, civil society organisations, journalists and ordinary citizens will therefore be operating in an information environment where distinguishing authentic material from synthetic content could become increasingly difficult.

One of the more serious possibilities discussed by experts is what has been described as “flooding the zone”. Instead of attempting to create one highly convincing false story, operators could generate thousands of pieces of content and distribute them through numerous websites and social media accounts. The objective would be to overwhelm the information environment rather than persuade people with one particular narrative.

Such a strategy could be especially difficult to counter in areas where reliable independent information is scarce. If there is a shortage of credible reporting on a particular political issue, a relatively small amount of money could potentially be used to fill the online space with large quantities of synthetic material.

There is also concern that this type of activity could eventually affect the information ecosystems used by artificial intelligence systems themselves. If enormous amounts of misleading content are placed online and repeatedly reproduced, future AI models could encounter distorted information when processing the digital record.

Anthropic has said that the political influence operations it identified were disrupted and that information from those investigations was used to strengthen its safeguards. The company has also said it shared intelligence with authorities and industry partners where appropriate.

The challenge for south-east Asian democracies will therefore extend beyond identifying individual deepfakes or fake social media accounts. Governments, technology companies, journalists and civil society groups may increasingly need to monitor coordinated networks, improve public awareness, strengthen verification systems and develop clearer safeguards around the use of AI in political communication.

At the same time, AI should not be viewed solely through the lens of electoral manipulation. The technology can provide legitimate benefits for journalism, research, translation, public communication and civic engagement. The risk arises when generative capabilities are combined with deceptive identities, personal data, coordinated distribution and deliberate attempts to exploit social divisions.

The Malaysian case offers an early illustration of how these elements can come together. It also demonstrates why the impact of AI on elections cannot be assessed simply by counting the number of fake images, videos or articles produced. The broader issue is how technology interacts with existing political tensions, social media networks and the public’s ability to distinguish reliable information from coordinated manipulation.

As south-east Asia moves towards another series of major elections, the region is entering a period in which political communication will increasingly coexist with rapidly evolving AI technology. The effectiveness of future safeguards will depend on cooperation between technology companies, election authorities, journalists, researchers and citizens.

The central challenge will be maintaining an information environment in which voters can encounter competing political ideas without being overwhelmed by artificially generated identities, fabricated evidence or coordinated deception. AI may change the speed and scale of political communication, but the resilience of electoral systems will continue to depend on the transparency of information, the strength of institutions and the ability of citizens to assess what they see and hear.