OpenAI introduces new ‘Trusted Contact’ safeguard for cases of possible self-harm
OpenAI Unveils Trusted Contact Feature for ChatGPT
OpenAI has announced a new safety feature named Trusted Contact for adult ChatGPT users. This tool is designed to notify a trusted friend or family member if someone expresses thoughts of self-harm while using ChatGPT.
How Trusted Contact Works
- Users can select a trusted individual to be alerted in cases where messages reference self-harm.
- If ChatGPT detects language related to self-harm, it encourages the user to reach out to their trusted contact.
- An automated alert is sent to the designated person via email, text, or an in-app notification, urging them to check in with the user.
- Alerts are intentionally brief and do not include detailed conversation content, helping to safeguard user privacy.
Background and Context
This move follows previous lawsuits against OpenAI, where families allege ChatGPT conversations led to tragedy. The company already uses both automated monitoring and human review to identify safety concerns, striving to address alerts within an hour.
The Trusted Contact feature expands on earlier safety measures like parental controls and professional health service referrals.
Optional and Privacy-Focused
- Enabling Trusted Contact is optional for users.
- Multiple ChatGPT accounts can exist for a single user, and similar options are present in parental controls.
OpenAI’s Commitment to Safety
OpenAI states this initiative is part of broader efforts to make AI helpful and supportive during critical situations, and pledges to continue working with experts to further refine these systems.
For more details, read the original article by Lucas Ropek at TechCrunch.
