AI firm Anthropic says it disrupted election manipulation operation targeting voters in Malaysia
Anthropic says its Claude AI was used to build a voter-targeting system covering all 222 parliamentary constituencies in Malaysia, while the threat actor also unsuccessfully pursued a contract with the country’s communications regulator.
Anthropic's threat intelligence report, published on Sep 10, 2026, includes a screenshot of the "Malaysia Pulse" website, which Anthropic said was a fabricated news site as part of the election manipulation operation.
This audio is generated by an AI tool.
JOHOR BAHRU: Artificial intelligence (AI) company Anthropic said it has disrupted a commercial election manipulation platform that primarily targeted people in Malaysia.
The operation involved around 1,000 fake social media accounts, a bogus news site and fabricated intelligence dossiers.
It used Anthropic’s Claude AI models to build a system capable of profiling and targeting Malaysian voters by constituency, the American company said in a threat intelligence report published on Thursday (Sep 10).
The report covers activity Anthropic said it disrupted between December 2025 and August 2026, including cyber operations, influence operations, surveillance, scams and fraud and other forms of misuse.
“The platform posed as a defensive cyber intelligence and counter-disinformation tooling outlet,” Anthropic said.
“Nevertheless, our investigation uncovered clear links to BBS Bilisim Teknolojileri, an Istanbul-based technology company, which sold access to the platform as a paid influence-as-a-service capability.”
Anthropic did not identify who paid to use the platform for the Malaysia-focused activity, or whether it was connected to any particular Malaysian election.
It also did not specify exactly when the Malaysia-focused operation took place or how long it ran.
In its case study titled "Disrupting a commercial election-manipulation platform targeting Malaysia", Anthropic said it identified and removed the Claude account involved and used information gathered from the account to map the wider operation.
CNA has contacted Malaysia's Communications Minister Fahmi Fadzil and the Malaysian Communications and Multimedia Commission (MCMC) for comment on Anthropic’s findings, as well as whether Malaysian authorities are investigating the operation and whether BBS Bilisim Teknolojileri or its representatives pursued a contract with the government.
MCMC is Malaysia's national regulator for the communications and multimedia industry and falls under the Communications Ministry.
1,000 FAKE X ACCOUNTS WITH “WARM-UP” MECHANISMS
According to the report, the operation used real census and electoral data as well as "millions of voter records" to build a voter-targeting system covering all 222 parliamentary constituencies in Malaysia.
The system was designed to target what Anthropic described as some of Malaysia's most sensitive political and social fault lines – race, religion and royalty.
Claude was used to build constituency profiles using the electoral data and to help develop the infrastructure underpinning the influence operation, Anthropic said.
The platform managed more than 1,000 fake accounts on social media platform X. According to the report, the accounts were equipped with "warm-up" mechanisms intended to make them appear authentic before being deployed in influence operations.
These included techniques such as regularly renewing cookies and IP addresses in an attempt to evade social media platforms' detection systems.
Anthropic said the platform's dashboard allowed its user to determine how many artificial views a particular target should receive.
In one instance, Anthropic said it observed a request "in support of the sitting Malaysian Prime Minister" for one million artificial views on his account.
The report did not identify who requested the artificial views or establish any connection between the operation and Prime Minister Anwar Ibrahim or his office.
Anthropic also cautioned that engagement figures recorded by the platform's own tools were self-reported by the actor and could not be independently verified.
It said one example involving the account of a senior Malaysian government official recorded engagement figures in the millions.
BOGUS NEWS SITE, FABRICATED DOSSIERS
The operation extended beyond social media.
Anthropic said the setup ran a fabricated news site called "Malaysia Pulse", using AI to scrape legitimate Malaysian news reports and repeatedly rewrite them before publishing the resulting articles under fabricated bylines.
The report said Malaysia Pulse also republished material from Chinese and Russian state-aligned outlets, including Xinhua, CGTN, Sputnik/RIA and TV BRICS.
According to Anthropic, references identifying those organisations as state-aligned sources were removed, allowing the material to be presented as independent Malaysian reporting.
Anthropic identified malaysiapulse.com as an actor-controlled domain and said it was registered on May 10 this year.
It also linked a YouTube channel, @malaysiapulseof, to the synthetic news operation.
At the same time, the people behind the operation used Claude to produce fabricated intelligence reports to spread false allegations against an opposition politician and civil society organisations, Anthropic said, without naming these parties.
“These allegations were entirely made up by the actors," Anthropic said.
CNA has reached out to Anthropic for details on the opposition politician and civil society organisations targeted by the false allegations.
Anthropic said its AI system refused or partially refused some of the requests made by the operator, including after Claude identified one dossier as material intended for political defamation.
The actor subsequently changed the wording of requests in an effort to continue building the same capabilities, the report said.
Anthropic said the actor behind the operation had pursued a contract with Malaysia's national communications regulator.
“We found no evidence that this pursuit succeeded,” the report said.
The report did not provide further details on the proposed contract, including when the approach was made, what services were offered or how far any discussions progressed.
NO EVIDENCE OPERATION REACHED AUTHENTIC COMMUNITIES
Despite the scale and sophistication of the infrastructure, Anthropic said it found no evidence that the campaign broke into authentic online communities.
“The assets were distributed across multiple platforms, but without evidence of breakout into authentic communities,” Anthropic said.
Anthropic said it detected the account through its internal systems, removed it and used indicators recovered during its investigation to map the operation's wider footprint and disrupt future misuse.
The Malaysia case was among several examples detailed in Anthropic's September threat intelligence report examining malicious use of its Claude AI models.
Other cases included a suspected Russia-linked cyber espionage campaign and efforts by Chinese AI firms it accused of trying to extract and replicate Claude's capabilities.
New media expert Saifuddin Ahmed from Nanyang Technological University’s Wee Kim Wee School of Communication and Information told CNA that the Malaysia operation outlined by Anthropic went beyond normal political campaigning and “crossed the line into manipulation” because it relied on deception and disguise.
“Instead of communicating openly, the operators reportedly created fake social media accounts designed to behave like real people and established a fake news website that republished legitimate stories,” said Saifuddin.
“It was therefore not simply a matter of targeting voters more effectively; it involved manufacturing an artificial, manipulated public conversation while concealing who was behind it,” he added.
He added that similar tools have reportedly been used to manipulate voters, citing how an Israeli group nicknamed Team Jorge allegedly used social media bots to manipulate elections in Africa and Latin America while a Russian-linked network known as Doppleganger has purportedly imitated legitimate news organisations and websites in Europe in order to circulate misleading or false political narratives.
“These cases show that influence operations are increasingly offered as a commercial service … and political manipulation can potentially be purchased by a wide range of actors,” he added.
