Anthropic's Claude AI Used in Malaysian Election Influence Operation

AI Used in Malaysian Election Disinformation

Anthropic recently disclosed that its Claude AI was utilized in a political influence operation during recent state elections in Malaysia. The operation involved creating a commercial election manipulation platform that leveraged approximately 1,000 X accounts, a fake news outlet named “Malaysia Pulse,” and a series of fabricated documents. This setup was designed to spread disinformation en masse.

Targeting Voters and Exploiting Divisions

The influence campaign appeared to target voters across all 222 Malaysian parliamentary constituencies. It reportedly used census and electoral data to exploit sensitive political and social divisions, including issues related to race, religion, and royalty. Although Anthropic intervened and the experiment was ultimately ineffective, it raised concerns about the potential weaponization of AI in future elections.

AI's Role in Scaling Disinformation

Katie Harbath, former head of public policy for global elections at Meta, noted that AI introduces a new level of sophistication and ease to the disinformation landscape. AI enables the creation of more personalized content, quicker and cheaper, effectively 'supercharging' disinformation efforts. However, distribution mechanisms, primarily social media platforms, remain crucial for these campaigns.

Broader Implications for Southeast Asia

Anthropic's September threat assessment detailed this Malaysian operation, among others, where actors used Claude to build networks of fake social media profiles and news websites targeting audiences globally. The report identified actors including governments, state-aligned propaganda institutions, state media, and private firms selling influence. Southeast Asia is particularly vulnerable to disinformation, with AI making such operations significantly easier by mimicking traditional 'troll factories'.