OpenAI announced an upgrade to ChatGPT’s restrictions on violent content. Frequent dangerous conversations will result in account bans. OpenAI recently officially announced that it will comprehensively upgrade and strengthen ChatGPT’s restrictions on violence-related content in the next few weeks. At the same time, it will launch a more stringent weapon and security system information protection mechanism for minor users.

OpenAI announced an upgrade to ChatGPT's restrictions on violent content. Frequent dangerous conversations will result in account bans. OpenAI recently officially announced that it will comprehensively upgrade and strengthen ChatGPT's restrictions on violence-related content in the next few weeks. At the same time, it will launch a more stringent weapon and security system information protection mechanism for minor users. The core of this policy adjustment is to adopt more decisive intervention methods in high-risk dialogues. Under the newly announced measures, ChatGPT will have the ability to proactively end conversations if the system detects and evaluates a user's conversations as showing a continued tendency to focus on ideas of violence or serious harm to others. What’s even more severe is that if a user is terminated by the system multiple times due to such issues, their account will face penalties of being suspended or even directly banned. At the specific content filtering boundary, the original policy prohibits providing any operational guidance to help users plan or commit violent acts. In this adjustment, the scope of restrictions has been further expanded to the historical level - the details of the model's answers to questions related to "past large-scale attacks and their perpetrators" will be greatly reduced, or even directly refused to answer. The specific restricted content not only covers the weapons and equipment involved in recreating the attack process and how to obtain them, but also restricts user queries for detailed comparisons of the scale of damage designed to amplify the attacker's visibility. It should be noted that this series of restrictions applies to all users, regardless of age group or question purpose. Even adult users will be subject to the same constraints when making relevant inquiries. In addition to the general restrictions for all users, OpenAI further adds refined protection for youth groups identified as under 18 years of age. The system will specifically restrict useful information about weapons and security systems that could be used to commit violence, and this protection will be automatically triggered even if the question does not appear to directly indicate violent intent. However, OpenAI emphasized that teenagers can still seek help through ChatGPT normally, such as learning how to prevent or mitigate violence, or getting effective guidance to connect with trusted adults. In order to ensure the accurate implementation of the youth protection mechanism, OpenAI has introduced an improved age prediction mechanism. This mechanism can combine multi-dimensional signals such as the user's daily behavior to continuously conduct a dynamic assessment of whether the account belongs to a minor, and automatically match the corresponding protection measures at the appropriate time. If the user's account is unfortunately misjudged as a minor by the system, it can also be verified and adjusted by actively submitting identity information. via AI News (author: AI Base)