OpenAI and Meta Strengthen AI Safety Safeguards for Teens
AI 안전 | Fri Jul 17 2026 00:00:00 GMT+0000 (Coordinated Universal Time) | 2 sources
OpenAI and Meta introduced AI safety features and parental control tools targeting teenage users.
Analysis
[OpenAI] strengthened ChatGPT safeguards and learning features for teens [1]
- Introduced age prediction functionality
- Expanded Parental Controls
- Enabled parents to directly activate Study Mode
- Provided education-focused starter prompts
[Meta] introduced parental alerts triggered by teen crisis signals [2]
- Built AI system to detect suicide/self-harm-related conversations
- Flagged conversations undergo manual review before alerts
- Errs on the side of alerting parents in ambiguous cases
- Rolled out first in US
- UK
- Australia
- and Canada
[Meta AI] developed emergency services connection feature for crisis situations [2]
- Detects imminent risk for both adults and teens
- Connects to emergency services in cases of imminent suicide/self-harm risk
- Developed risk signals in collaboration with experts
[OpenAI Safety Principles] announced four principles for teen safety [1]
- Prioritize teen safety even when it conflicts with other goals
- Encourage real-world support when needed
- Treat teens as teens
- Ensure transparency through clear expectations