OpenAI has introduced an image classifier to detect child sexual abuse material, including AI-generated images.
The tool is available free to selected companies through an API; in initial tests, it reduced by 48% the images requiring human review.
OpenAI has also published developer guidance on detection and reporting.
Detection tools and rules in ChatGPT
OpenAI has created a new image recognition system, available for free to select partner companies. The tool analyzes images to identify visual signs of abuse. In initial tests, the system increased detection accuracy and reduced the number of images requiring human review by 48%.
At the same time, ChatGPT is getting warning screens for users who search for content that sexualizes children. When such an attempt is detected, the system will display a warning and direct the user to support services. In addition, the user will be banned from the service for 24 hours. If someone continues to violate these rules, their account may be permanently deleted.
Collaboration with experts and authorities
The new measures were developed in collaboration with organizations including NCMEC, Tech Coalition, and ROOST. OpenAI has also published guidance for developers on how to identify and report such material. It is also providing specialized tools and training to law enforcement agencies to help them handle cases involving the safety of minors.
>Stay connected with Hacks.gr
Set us as a preferred source on Google and find us on social media.
Preferred source
↗
end of article
Top comments (0)