OpenAI and Anthropic will start predicting when users are underage

OpenAI and Anthropic are updating their AI chatbots with enhanced safety protocols, particularly for users aged 13 to 17. OpenAI has updated ChatGPT's Model Spec with four new principles prioritizing teen safety, even over goals like maximum intellectual freedom, and encouraging real-world support. This follows increased scrutiny over AI's mental health impact, including a lawsuit against OpenAI regarding self-harm instructions. OpenAI is also developing an age prediction model to automatically apply teen safeguards if a user appears to be under 18. Similarly, Anthropic is developing a system to detect subtle conversational signs indicating a user might be underage during interactions with Claude, disabling accounts confirmed to belong to minors. Anthropic also detailed progress in reducing sycophancy in its models, noting that while improvements are being made, there is a trade-off between model warmth and reducing harmful affirmation.

(Source：The Verge)