Anthropic Says Claude Was Used to Run Over 20 AI Dating Apps

Anthropic released a new AI abuse report saying app development studios used Claude to build and operate more than 20 dating apps. According to ChainCatcher, about 75% of the match profiles shown to ordinary users were AI characters, while the platforms claimed all users were real people. During a two-week observation period in April 2026, the apps ran more than 4,700 AI characters, chatted with at least 25,000 different users, and sent about 2.36 million messages. Most of the affected users were in the United States. The platforms reportedly mixed profiles at a ratio of about three AI characters for every one real person. Claude maintained thousands of conversations at the same time, while part-time human workers stepped in when users requested video calls, photos, or social media follows. The system generated three suggested replies for each message, and users had to spend credits to chat and match before buying in-app virtual currency to continue. Fully AI-controlled accounts were instructed not to admit they were bots. To avoid app store review, operators reportedly used special interfaces that were only enabled during review periods and different program class names across more than 20 apps. Anthropic said the related brands included DORA, DONI, ROMI, LUMA, JOVIA, KIRA, GRACECHAT, HAVEN, NALO, and LOVIA. The company said it has banned the related accounts and shared the app identities and review-evasion details with Apple and Google.