← Back to live feed1 story
A studio based in China used Claude to run 4,700 AI personas across more than 20 fake dating apps, Anthropic said. The personas chatted with 25,000 users and generated 2.36 million messages over two weeks.
Anthropic published a threat intelligence report on Sept. 10 detailing efforts to misuse Claude. The company said it disrupted every operation covered in the report and used its findings to strengthen safeguards, sharing information with authorities and other AI companies where appropriate. It cautioned that the cases were not typical uses of Claude.
Sign in to suggest edits
Key sources
- SUPPORT@coinmarketcap“Anthropic says it caught a China-based studio using Claude to power 4,700 AI personas across 20+ fake dating apps, chatting with 25,000 users and generating 2.36M messages over two weeks.”x.com
- SOURCE@afp“Anthropic says it has shut down multiple cases of state-sponsored surveillance operations that used its AI models.”x.com
- SUPPORT@cnbc“Anthropic said it detected and disrupted unauthorized large-scale efforts by China-based AI labs including Alibaba, Moonshot and DeepSeek to train their models using Claude.”x.com
- SUPPORT@clashreport“Anthropic says Moonshot generated 23M+ Claude exchanges from May–July 2026 through 5,380 allegedly fraudulent accounts.”x.com
- SUPPORT@clashreport“Anthropic says there's no evidence they fielded an operational weapon — but they already had an offline toolkit that no longer needed Claude.”x.com
- SUPPORT@marcowenjones“Anthropic's latest report on detecting misuse of AI contains this case of an anti-Muslim Brotherhood campaign by the UAE. It includes a sockpuppet network of 300 accounts.”x.com
- SUPPORT@marionawfal“Anthropic tied other accounts to Iranian state-aligned institutions running "soft war" and "cognitive warfare" campaigns, using Claude to build fake personas, target databases, campaign plans and social-media content in 20 languages.”x.com
- SUPPORT@lukolejnik“A Russian espionage group had AI agents check if security tools had caught its malware, then rebuild it until they hadn't.”x.com