Anthropic has revealed details of a political influence operation in Malaysia that used its Claude AI to create a commercial election manipulation platform. The operation involved a network of about 1,000 X accounts, a fake news outlet named “Malaysia Pulse”, and a series of fabricated documents.
The operation appeared to target voters in recent state elections, using census and electoral data to exploit sensitive political and social divisions across all 222 Malaysian parliamentary constituencies. These divisions included race, religion, and royalty.
Anthropic's latest threat assessment, released in September, detailed how actors used Claude to build networks of fake social media profiles and news websites that targeted audiences on six continents. The report stated that these actors included “governments, state-aligned propaganda institutions and state media, as well as private firms selling influence”.
Katie Harbath, former public policy lead for global elections at Meta, noted that AI brings a new level of sophistication and ease to disinformation, allowing for more personalised content to be made quicker and cheaper. Rossine Fallorina of the Philippine thinktank Sigla added that AI speeds up the production process, mimicking what was previously termed a “troll factory”.
Anthropic stated that in each identified case of political influence, the company “disrupted the activity, used what we learned to strengthen our safeguards, and shared intelligence with authorities and industry partners, where appropriate”.