AI Tools Empower Political Disinformation Campaigns Across Southeast Asia
Experts warn that artificial intelligence is empowering political influence operations in Southeast Asia, as revealed by Anthropic's exposure of a disinformation campaign utilizing Claude in Malaysia.
The dangers of artificial intelligence are gaining global attention, as experts warn that revolutionary technology could provide political influence operators across Southeast Asia with powerful new tools for spreading disinformation. This is particularly concerning in a region characterized by young populations heavily reliant on social media.
One alarming example was recently exposed by Anthropic, which detailed how its Claude AI model was exploited in a political influence campaign in Malaysia. The operation involved building a commercial platform designed for election manipulation, utilizing a network of approximately 1,000 accounts on X, a fake news site called "Malaysia Pulse," and a series of fabricated documents.
According to Anthropic, the campaign was likely aimed at swaying voters during recent political cycles in Malaysia. It leveraged census and election data to exploit sensitive societal divisions—such as race, religion, and the monarchy—across all 222 parliamentary constituencies.
While the operation was thwarted after Anthropic intervened, its exposure has raised alarms about how AI might be weaponized in upcoming elections across Southeast Asia, a region already grappling with widespread fake news and disinformation.
Katie Harbath, a former public policy director at Meta, noted that the Malaysia case illustrates how AI transforms disinformation. She explained that AI enables the rapid, low-cost production of customized content at unprecedented scales.
"Artificial intelligence can supercharge such events—it enables the creation of more content, personalized, faster, and cheaper," said Harbath, who currently leads the tech advisory firm Anchor Change. However, she emphasized that a distribution mechanism, typically provided by social media platforms, is still required.
Anthropic detailed the Malaysia activity in its latest report published in September. The report described how various actors used Claude to establish networks of fake social media profiles and fictitious news outlets targeting audiences across six continents.
The company stated that actors involved in such influence operations include state governments, state-backed propaganda bodies, state-run media, and private firms selling influence services.
Southeast Asia is no stranger to disinformation. In recent years, Filipino journalist and Nobel Peace Prize laureate Maria Ressa compared the Philippines to a "petri dish" for the abuse of disinformation and fake news—techniques tested locally before being successfully replicated in the West.
However, artificial intelligence threatens to make the process vastly simpler.
"Basically, AI simply mimics what a few years ago we would have called a troll factory," said Rossine Fallorina of the Philippine research institute Sigla. He noted that the speed of generation removes traditional bottlenecks: "You do not need ten people to craft a narrative or a story. You simply write a prompt and distribute it."
Last year, OpenAI blocked accounts that used artificial intelligence to generate comments on Philippine politics and public figures across social media. Many of the fake comments expressed support for President Ferdinand Marcos Jr.
Experts fear that the deployment of AI on social media could exacerbate political and social fractures. "It feeds political polarization and preys on existing divisions," said Bridget Welsh, an associate professor at the University of Nottingham Malaysia.
General elections in Malaysia are scheduled for late 2027, while the Philippines and Cambodia anticipate elections in 2028, followed by Indonesia the subsequent year.
Dr. Ben Loh, a communications lecturer at Monash University Malaysia, noted that AI-generated content has been used to stoke hatred against Rohingya refugees who fled Myanmar for Malaysia amid the civil war.
"People already arrive with this deep-seated hatred rooted inside them, and when they see such content, even if it is generated using AI, it simply reinforces what they already believe to be true," he said.
Anthropic's report revealed that the fake news site "Malaysia Pulse" copied genuine reports from Malaysian media outlets, then used the AI model to repeatedly rewrite them and republish them under fictitious journalist bylines.
The site also republished articles from foreign state-linked media outlets associated with Russia and China, including TV BRICS, Xinhua, Sputnik/RIA, and CGTN, while stripping away attribution to the original sources.
"In each case where political influence activity was identified, the company disrupted the activity, used what it learned to strengthen its defenses, and shared intelligence with authorities and industry partners as appropriate," the report concluded.
