Beta The Briev beta is out. Free on iPhone via TestFlight — install it in under a minute.

Join the beta ↗
Briev
Live
Politics
CROSS-SPECTRUMBROAD COVERAGE

Anthropic disrupts AI-driven platform that used fake accounts to sway Malaysian elections

Anthropic says it shut down a commercial operation that employed its Claude AI to run thousands of fake social-media accounts and a bogus news site aimed at influencing Malaysia’s election.

Anthropic’s report details a covert campaign that leveraged its Claude model to power a commercial election-manipulation service focused on Malaysia. The operation ran roughly 1,000 fake X accounts, constantly rotating cookies and IP addresses to evade detection, and operated a synthetic news site named Malaysia Pulse that rewrote genuine Malaysian stories and republished content from Chinese and Russian state media without attribution.

Using census and electoral data, the actors profiled voters constituency by constituency, exploiting fault lines such as race, religion and the monarchy. The group marketed the system as a "military-grade, AI-driven, real-time political operations ecosystem" and even sought a contract with the national communications regulator, though no evidence of success was found. Anthropic says Claude declined several requests, including the creation of fabricated defamation dossiers, and rates the campaign as a Category Two breakout on its scale.

Why it matters

It shows how advanced AI can be weaponized to manipulate elections, raising urgent governance and security concerns.

How the sides frame it

MODERATE AGREEMENT

Left-leaning coverage emphasizes the covert AI-driven manipulation operation itself, while centrist and right-leaning coverage focus on the Malaysian regulator’s response and denial of involvement.

LEFT

Highlights the scale and tactics of the AI-powered election-manipulation campaign targeting Malaysia.

CENTER

Reports that the regulator takes the Anthropic findings seriously but says the issues fall outside its mandate.

RIGHT

Frames the regulator’s statement as a rejection of any involvement in the alleged AI-driven meddling.

The left emphasises

  • the operation used Anthropic’s Claude model and ran about 1,000 fake X accounts
  • a synthetic news site rewrote Malaysian stories and republished Chinese and Russian state media without attribution
  • actors profiled voters constituency by constituency, exploiting fault lines such as race, religion and the monarchy

The right emphasises

  • MCMC asserts the described activities are not part of its functions
  • the commission says it is aware of the report and is reviewing the information
  • any action will be based on evidence after seeking clarification

In this story

AI election manipulationClaudefake social media accountsdisinformationvoter profilingMalaysia Pulsethreat intelligence
Get the beta ↗