Overview

• Meta said it is retraining its AI so it does not engage adolescents on self-harm, suicide, eating disorders or romantic/sexual topics and will instead direct them to specialized resources.
• The company will temporarily limit teens’ access to certain character bots, allowing only education- and creativity-focused chatbots, and described these steps as provisional.
• A Reuters probe found dozens of chatbots on Meta’s platforms that impersonated celebrities without consent, some generating sexualized content and images, including of 16-year-old actor Walker Scobell.
• Meta removed about a dozen of the impersonator bots and reiterated that its policies prohibit nude, intimate or sexually suggestive images, according to spokesperson Andy Stone.
• OpenAI said conversations indicating plans to harm others are routed to a restricted human-review flow and, if an imminent threat is confirmed, may be referred to law enforcement, a process drawing privacy scrutiny alongside lawsuits and a 44‑state attorneys‑general warning over child safety.