Sex-Fantasy Chatbots Are Leaking a Constant Stream of Explicit Messages

UpGuard collected roughly 1,000 prompts exposed by AI instances over a 24-hour period; the dataset included multilingual prompts and several sexually explicit role-playing scenarios, some allegedly involving children. The researcher could not identify a specific leaking service and noted the likely origin as small or individual AI model instances; no usernames or direct personal identifiers were present in the collected prompts.