Overview

• Accounts for minors must be linked to an adult before protections apply, with options to disable memory, voice and image generation and to set off‑hours.
• Sexual, romantic or violent roleplays and other inappropriate content, including viral challenges and extreme beauty ideals, are now blocked.
• Parents will receive notifications when serious risks are suspected, but they will not get access to their children's full chat histories.
• ChatGPT will silently route sensitive or emotional queries to a reasoning model or a GPT‑5 “chat‑safety” variant, though OpenAI has not detailed what triggers the switch.
• The measures are available now, including in Germany, after OpenAI acknowledged prior hotline‑style safeguards could fail during long conversations.