Meta Platforms is testing a new breed of generative-AI chatbots that can initiate conversations with users across Facebook Messenger, WhatsApp and Instagram, according to leaked internal documents reported by Business Insider and confirmed by the company to TechCrunch. The initiative, codenamed “Project Omni,” is being developed with data-labeling firm Alignerr Corp. and relies on Meta’s AI Studio, which lets customers build their own virtual personas. Under current guidelines, a bot may send one unsolicited follow-up message—but only to people who have exchanged at least five messages with it during the previous 14 days—and must cease contact if the user does not respond. Meta says the proactive bots are intended to improve re-engagement and brand retention, while steering clear of sensitive topics unless a user raises them first. The company has issued disclaimers that the chatbots are not substitutes for professional advice and may produce inaccurate or inappropriate responses. The experiment is part of a broader effort to monetise AI products. Court filings show Meta projects generative-AI revenue of $2-3 billion next year and as much as $1.4 trillion by 2035. The push also sharpens scrutiny of user safety and privacy, echoing controversies faced by rival Character.AI, which is being sued after allegations that one of its chatbots contributed to a teenager’s death.