Anthropic, a U.S.-based artificial intelligence company, has revealed that it disrupted an AI-powered platform designed to manipulate public opinion during an election in Malaysia. The operation reportedly involved over 1,000 fake accounts on the social media platform X (formerly Twitter), a fabricated news website, and the creation of false intelligence reports aimed at influencing voters. The system targeted all 222 parliamentary constituencies in Malaysia, using millions of voter records to identify and exploit political and social divisions, such as race, religion, and the monarchy. The platform, which posed as a legitimate cyber intelligence and counter-disinformation tool, used AI to generate synthetic content and manage a network of fake accounts. These accounts were designed to mimic real users by frequently changing their digital fingerprints, such as cookies and IP addresses, to avoid detection. According to internal documentation from the threat actors, the infrastructure was described as a "military-grade, AI-driven, real-time political operations ecosystem." The campaign also operated a fake news website called "Malaysia Pulse," which used AI to rewrite real news articles and publish them under invented names. It also incorporated content from Chinese and Russian state-linked media, removing any direct references to those sources. Claude, Anthropic's AI model, was used to create fabricated intelligence reports that included false claims about an opposition politician and civil society groups. These allegations were entirely made up by the actors. Anthropic classified this campaign as a "Category Two" on its Breakout Scale, which indicates that the assets were spread across multiple platforms but did not successfully infiltrate real communities. The company also noted that the operator had attempted to secure a contract with Malaysia’s national communications regulator, but found no evidence that this effort was successful. Anthropic reported that Claude either refused or partially refused the actor's requests at several points, including when it detected that a fabricated dossier was intended for political defamation or when it encountered language that suggested a psychological operation. The company did not confirm any direct link between the operation and Malaysian Prime Minister Anwar Ibrahim or his office, though it observed a request for "one million artificial views" on the prime minister's account. The exact source of this request remains unclear.