OpenAI expands ChatGPT safety measures, targeting historical violence and minor users

OpenAI is implementing significant updates to ChatGPT's content restrictions, moving beyond the core announcement of proactive conversation termination for high-risk discussions. The enhancements introduce stricter limitations on historical content, specifically reducing or outright refusing responses to queries concerning past mass attacks and their perpetrators. Furthermore, the scope of restrictions is broadened to include limiting details on weapons, equipment, and methods for acquiring them related to recreating attacks, and also restricts detailed comparisons intended to amplify an attacker's notoriety. These new constraints apply universally across all user demographics. For users under 18, protections are intensified, automatically restricting practical information on weapons and security systems even without explicit violent intent. To manage this, OpenAI has integrated an improved age-prediction mechanism that dynamically assesses user behavior, allowing for identity verification if a minor is misclassified.