Anthropic Disrupts AI Election Manipulation Platform
AI firm Anthropic says it disrupted a paid influence operation that used its Claude AI to build a voter-targeting system for all 222 Malaysian
Artificial intelligence company Anthropic says it disrupted a commercial election manipulation platform that primarily targeted voters in Malaysia. The operation, active between December 2025 and August 2026, used Anthropic's Claude AI models to build a system for profiling and targeting Malaysian voters by constituency, according to a threat intelligence report the company published on September 10.
Anthropic stated the platform posed as a defensive cyber intelligence and counter-disinformation tool. Its investigation, however, 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. The AI firm did not identify who paid for the Malaysia-focused activity or connect it to any specific election.
Voter Targeting and Fake Social Media Accounts
The operation used real census and electoral data alongside millions of voter records to build a voter-targeting system covering all 222 parliamentary constituencies in Malaysia. According to Anthropic's report, the system was designed to target sensitive political and social fault lines in the country, namely race, religion, and royalty. Claude AI was used to build constituency profiles and develop the infrastructure for the influence operation.
The platform managed more than 1,000 fake accounts on the social media platform X. These accounts were equipped with "warm-up" mechanisms intended to make them appear authentic before deployment. Techniques included regularly renewing cookies and IP addresses to evade detection systems. The platform's dashboard allowed its user to determine how many artificial views a target should receive.
Anthropic said it observed one request "in support of the sitting Malaysian Prime Minister" for one million artificial views on his account. The report did not identify who made the request or establish any connection to Prime Minister Anwar Ibrahim or his office. The company cautioned that engagement figures recorded by the platform's tools were self-reported and unverified, noting one example involving a senior Malaysian government official's account recorded engagement in the millions.
Fabricated News and Intelligence Dossiers
The operation extended beyond social media with a fabricated news site called "Malaysia Pulse." The site used AI to scrape legitimate Malaysian news reports, repeatedly rewrite them, and publish the articles under fabricated bylines. Malaysia Pulse also republished material from Chinese and Russian state-aligned outlets like Xinhua, CGTN, Sputnik/RIA, and TV BRICS, removing references identifying them as state-aligned to present the material as independent Malaysian reporting.
Anthropic identified malaysiapulse.com as an actor-controlled domain registered on May 10 this year and linked a YouTube channel, @malaysiapulseof, to the synthetic news operation. Also, the actors used Claude to produce fabricated intelligence reports to spread false allegations against an opposition politician and civil society organizations. "These allegations were entirely made up by the actors," Anthropic stated.
The company noted its AI system refused or partially refused some requests, including after Claude identified one dossier as material for political defamation. The actor subsequently changed the wording of requests to continue building the same capabilities.
Pursuit of Government Contract and Disruption
Anthropic's report revealed that the actor behind the operation had pursued a contract with Malaysia's national communications regulator, the Malaysian Communications and Multimedia Commission (MCMC). "We found no evidence that this pursuit succeeded," the report said, without providing further details on the proposed contract or discussions.
Despite the scale of the infrastructure, Anthropic said it found no evidence the campaign broke into authentic online communities. The company detected the account through its internal systems, removed it, and used indicators from its investigation to map the operation's wider footprint and disrupt future misuse. The Malaysia case was among several detailed in Anthropic's September threat report examining malicious use of its Claude AI models.
