Yes, yes, the AI will begin to drain information about your mood to third parties absolutely legally.
OpenAI began to introduce a new security feature for ChatGPT, which allows adult users to assign a trusted person for emergency cases. If the system notices signs of a serious risk of self-harm or suicide, the selected contact may be notified.
The feature is called Trusted Contact and is still working as a voluntary setting. Previously, similar mechanisms the company used only in parental control for adolescents. Now the tool has become available to adults over 18 years old, and in some countries - over 19.
The user can specify one trustee: a friend, relative, partner or guardian. After inviting, the contact must confirm the consent within a week, otherwise the adjustment is not activated.
OpenAI explains the scheme of work as follows: automatic ChatGPT systems analyze messages and look for signs of a serious crisis. If the algorithm notices a potentially dangerous situation, the user first receives a warning. ChatGPT offers to ask for help on your own and reports that the trusted contact can be notified.
After that, the case may be checked by a small group of OpenAI employees who have undergone special training. If the specialists consider the threat to be real, the trustee will be sent a short notification by e-mail, SMS or inside the application. ChatGPT correspondence will not be disclosed. In the notification will appear only a general description of the situation and recommendations on how to support a person.
The company says it has developed a function together with clinicians, researchers and the American Psychological Association. The launch of Trusted Contact takes place against the backdrop of a series of lawsuits against OpenAI. Relatives of users in several cases claimed that communication with chatbots intensified mental problems and pushed people to self-harm.
OpenAI has previously denied accusations that its systems encourage dangerous behavior, but pressures around the topic of AI security continue to grow. On Reddit, some users met a new feature cautiously positively. One commentator wrote that the idea can help people if the implementation turns out to be careful:
In one of the popular threads in X, they drew attention to the main controversial point: control over the system remains at the platform. OpenAI determines which messages are considered alarming, what cases to be transmitted to people for verification and when to notify a third party. At the same time, a trusted contact may be an ordinary relative or friend without psychological training, which in some situations will only make worse. Suppose, the parent, with their branded phrases in the spirit of "well, I said" or "it's all about idleness." Especially if the only contact...
Other critics have called Trusted Contact the next “security theater” and an attempt to show activity around the problem, which, in their opinion, is initially exaggerated. Some Reddit users doubt that the feature will help adults in a serious condition at all. In their opinion, a person with serious psychological problems will rather hide his condition and will not leave contact details in advance.
OpenAI emphasizes that the trusted contact can be removed or replaced at any time, and the person himself has the right to refuse to participate. The company also claims that the notifications specifically made as limited as possible in content so as not to turn the security function into a surveillance system.
OpenAI began to introduce a new security feature for ChatGPT, which allows adult users to assign a trusted person for emergency cases. If the system notices signs of a serious risk of self-harm or suicide, the selected contact may be notified.
The feature is called Trusted Contact and is still working as a voluntary setting. Previously, similar mechanisms the company used only in parental control for adolescents. Now the tool has become available to adults over 18 years old, and in some countries - over 19.
The user can specify one trustee: a friend, relative, partner or guardian. After inviting, the contact must confirm the consent within a week, otherwise the adjustment is not activated.
OpenAI explains the scheme of work as follows: automatic ChatGPT systems analyze messages and look for signs of a serious crisis. If the algorithm notices a potentially dangerous situation, the user first receives a warning. ChatGPT offers to ask for help on your own and reports that the trusted contact can be notified.
After that, the case may be checked by a small group of OpenAI employees who have undergone special training. If the specialists consider the threat to be real, the trustee will be sent a short notification by e-mail, SMS or inside the application. ChatGPT correspondence will not be disclosed. In the notification will appear only a general description of the situation and recommendations on how to support a person.
The company says it has developed a function together with clinicians, researchers and the American Psychological Association. The launch of Trusted Contact takes place against the backdrop of a series of lawsuits against OpenAI. Relatives of users in several cases claimed that communication with chatbots intensified mental problems and pushed people to self-harm.
OpenAI has previously denied accusations that its systems encourage dangerous behavior, but pressures around the topic of AI security continue to grow. On Reddit, some users met a new feature cautiously positively. One commentator wrote that the idea can help people if the implementation turns out to be careful:
But the first wave of discussions in social networks turned out to be mostly critical. Users began to argue not so much about the very idea of support, but about privacy and about who exactly will decide when the situation becomes dangerous.
In one of the popular threads in X, they drew attention to the main controversial point: control over the system remains at the platform. OpenAI determines which messages are considered alarming, what cases to be transmitted to people for verification and when to notify a third party. At the same time, a trusted contact may be an ordinary relative or friend without psychological training, which in some situations will only make worse. Suppose, the parent, with their branded phrases in the spirit of "well, I said" or "it's all about idleness." Especially if the only contact...
Other critics have called Trusted Contact the next “security theater” and an attempt to show activity around the problem, which, in their opinion, is initially exaggerated. Some Reddit users doubt that the feature will help adults in a serious condition at all. In their opinion, a person with serious psychological problems will rather hide his condition and will not leave contact details in advance.
OpenAI emphasizes that the trusted contact can be removed or replaced at any time, and the person himself has the right to refuse to participate. The company also claims that the notifications specifically made as limited as possible in content so as not to turn the security function into a surveillance system.
