Anthropic Exposes AI Dating Scam Involving 25,000 Users

ChainCatcher reports that Anthropic’s latest AI misuse report reveals that app development studios have used Claude to develop and operate over 20 dating apps. These platforms claim all users are real, but approximately 75% of the matches ordinary users encounter are AI-generated personas. During a two-week observation period in April 2026, these dating apps collectively ran over 4,700 AI personas, engaging in conversations with at least 25,000 unique users and sending approximately 2.36 million messages in total—primarily affecting users in the United States. The platforms mix profiles at a ratio of roughly three AI personas to one real person. Claude simultaneously maintains thousands of conversations, while human contractors intervene when users request video calls, photos, or social media follows. The system generates three optional replies for each message, allowing contractors to select one. User chats and matches consume credits, which must be replenished by purchasing in-app virtual currency. AI-controlled accounts are strictly prohibited from disclosing they are bots. To evade app store reviews, operators designed special interfaces that activate only during the review process; over 20 apps used different package names. Anthropic disclosed related brands including DORA, DONI, ROMI, LUMA, JOVIA, KIRA, GRACECHAT, HAVEN, NALO, and LOVIA. The company has suspended the associated accounts and provided details on app identities and review evasion techniques to Apple and Google.