Partager:

JAKARTA After being sued on charges of damaging the mental health of teens, Character AI decided to change things. This chatbot has changed the training model and added a new feature.

To build a safer space, Character AI says that they continue to build and fix large language models (LLMs) that interact directly with users. They even create separate models specifically for teens.

"The goal is to direct the model away from specific responses or interactions, reducing the chances of users finding, or encouraging models to restore, sensitive or suggestive content," Character AI said on its official website.

Character AI has also used two new tactics to prevent LLM from producing sensitive and inappropriate content. The platform seeks to block inappropriate content outputs and block inappropriate user inputs.

In the output section, the AI Charger uses a classification tool that can identify certain types of content and responses from LLM, while in the input section, the AI Charger improves filter detection and intervention capabilities on its platform.

"For example, if we detect that users have sent content that violates our Terms of Service or Community Guidelines, the content will be blocked from their conversations with Characters," Character AI explained.

In addition to preventing through improved models, Character AI has also launched a number of new features that can prevent mental health problems in adolescents. This feature is Parental Control to provide insight into their children.

Parents can see the time spent on the platform and characters that are often used to interact. In addition, there is also a Time Notification feature that is spent. Every time a user uses a platform for an hour, they will get a message.

"While this feature will later have customizable options for adult users, users under 18 will have more restrictions on their ability to change it," said Character AI.


The English, Chinese, Japanese, Arabic, and French versions are automatically generated by the AI. So there may still be inaccuracies in translating, please always see Indonesian as our main language. (system supported by DigitalSiber.id)