How the Trusted Contact Feature Functions
OpenAI has officially launched the Trusted Contact feature for adult ChatGPT users. This optional setting allows adding a contact person (family member or friend) who will receive a notification if AI algorithms detect signs of a serious mental health crisis or self-harm intentions. Unlike standard moderation systems that simply block content, this tool aims to engage real-world support from the user’s personal network.
The activation process begins in the account settings. The user provides the contact details of a person who must be at least 18 years old. Following this, the selected individual receives an invitation that they must accept within 7 days. It is important to note that the trusted contact does not need their own OpenAI account to receive notifications via SMS or email.
Multilevel Verification System and Privacy
To prevent false positives, OpenAI utilizes a hybrid approach. First, automated systems analyze the context of the dialogue. If critical markers are detected, the data is passed to a specially trained team of specialists for manual review. Only after a real threat is confirmed is a brief message sent to the trusted contact. The developers emphasize that the notification text does not include chat logs or conversation details to maintain the user’s basic privacy level.
Ethical and Legal Aspects of Implementation
The launch of Trusted Contact comes amid growing pressure on AI developers from regulators and the public. Recent lawsuits have pointed out that large language models may inadvertently reinforce destructive states in users. The new feature is an attempt by OpenAI to demonstrate a responsible approach to safety. However, experts note that this also places a certain responsibility on the trusted contact, who becomes the first line of support.
Economic Context and Availability
Implementing such complex systems requires significant investment in manual review teams. OpenAI estimates the cost of maintaining the safety infrastructure at millions of dollars annually. Currently, the feature is available for personal accounts and does not extend to business versions like ChatGPT Enterprise or Edu. The company plans a gradual rollout across different regions considering local data protection laws.
Future Development of Safety Systems
Trusted Contact is part of a broader strategy that includes updates to the Model Spec and the introduction of parental controls for teenagers. OpenAI is also collaborating with the American Psychological Association to refine algorithms for recognizing subtle linguistic markers of depressive states. It is expected that in the future, similar features will become standard for all major generative AI platforms.
0 Comments