Tech • IA • Crypto
ChatGPT alerts your loved ones
7/10TL;DR
OpenAI is rolling out a “Trusted Contact” feature that alerts a designated person if an AI detects signs of self-harm risk in a user’s conversations.
KEY POINTS
Launch of Trusted Contact
OpenAI has begun deploying a new feature called Trusted Contact in the United States, with plans for broader expansion. The tool allows users to designate a single trusted individual who can be notified if concerning behavior is detected. The feature is positioned as an added safety layer within personal AI use.
How the system works
Once activated, the system analyzes user conversations to identify signals of distress, including self-harm ideation or severe emotional downturns. Users retain control and can enable or disable this monitoring at any time. The analysis operates continuously while the feature is active.
Escalation process
If risk indicators are detected, the AI first responds directly to the user with calming prompts and suggestions to seek professional help, such as medical or psychological services. In more serious cases, the interaction may be escalated to a human reviewer within OpenAI for further assessment.
Notification of trusted contact
When escalation thresholds are met, the designated contact receives an alert indicating that the user may need support. The notification does not include conversation details but encourages the contact to check in directly with the user. This preserves a level of privacy while enabling intervention.
Human oversight and responsibility
The inclusion of human reviewers introduces an additional layer of judgment before alerts are sent. This hybrid approach reflects a broader effort to combine automated detection with human decision-making in sensitive situations.
Privacy and ethical concerns
The feature raises questions about data monitoring, consent, and the extent of AI oversight in private conversations. Some observers suggest it could shift responsibility away from the platform by involving third parties, while others view it as a proactive safety measure.
Availability and limitations
Trusted Contact is expected to be available across most individual subscription tiers but excluded from team or enterprise environments, where privacy boundaries differ. Adoption will likely depend on user comfort with sharing signals of vulnerability.
CONCLUSION
The Trusted Contact feature marks a significant step toward integrating human support into AI interactions, balancing safety ambitions with ongoing concerns about privacy and responsibility.