Platform fight with private chats
Platform operators like for example Microsoft, Facebook and Google are now allowed to scan private chats between users for child pornography. That includes of course the exchange of pictures of abused children and teens. The parliament has passed a law with 537 MPs, who voted for the so-called chat control initiative and a quarter was against or abstented.
Providers are already scanning messages across the board to prevent child pornography as well as for signs of grooming, i. e. the stalking of adults on minors, regardless of suspicion. Facebook however has paused the scanning and wishes to continue as soon as the decision has been made.
The question which now arises is: how exactly shall this task be done in an affordable way?
Due to the huge amounts of data on well-known platforms, the use of humans to check the image, video or text material to prevent child pornography is actually not a serious option.
The best available technology to prevent child pornography therefore would be artificial intelligence, which can be configured to work within data protection legislation.
airis:protect chat moderation solution
airis:protect  is providing an AI technology, which is capable to qualify and quantify persons in a picture. It is capable to identify for example the sex and age group (senior, adult, child) of a person, nudity and single body characteristics (face, hand, feet etc.). It is also capable to identify whether someone is having sex currently. Furthermore, there are plenty of additional rules implemented and are ready to be used at any time. This way the technology of airis:protect can be used to prevent child pornography.
However, ist is important to know that airis:protect has not trained the AI to child porn, due to the sensitive material which would be necessary for the highly specialised training. This  would be forbidden at least by European law to own such images or videos. Without the involvement of a governmental unit, it is not possible to use an AI for this task.
Currently, airis:protect is working to start a collaboration with a governmental unit, which would be allowed to own and therefore train the AI by themselves. The training itself is not capable to produce images in any way.
Final words
Last but not least, the whole system is working along German data protection law and the EU GDPR. The technology of irisnet would be the best option to counter the task arising from the opportunity to scan user to user communications according the law in the EU.
German speaking links:
