In an alarming report, Human Rights Watch revealed that several artificial intelligence tools were trained withs of children retrieved from the Internet without consent. Four years after they were posted online.

Behind the almost infinite possibilities of artificial intelligence, there is a common functioning of all the tools that use it. AI models must "train" with maximum data. One suspects that the Internet is a goldmine for this. More and more services and companies are no longer hiding their ambitions: what you post on the web is likely to feed an AI. Slack is accused of doing so with your messages, for example. For its part, Google is using what you write on the community social network Reddit, while Abode has been in the news by suggesting that your creations would serve to drive its in-house artificial intelligence.

>>>>>>Replacement 3500mAh NB-35 for South NTS-360RM

In some cases, this kind of practice even pushes associations militating for the respect of the private life of Internet users to file a complaint in several European countries. But if in the cases cited here, the intentions of the firms are quite clear, it becomes much more problematic when the data used by AIs are collected without consent. Especially when they concern populations more at risk like children. This is what Human Rights Watch discovered, and as far as I can tell right away, his findings have something to make parents shudder.

The analysis focuses on LIAON-5B. It is a database created by the non-profit organization LAION in Germany. Available for free and available to everyone, it currently contains 5.85 billion photos associated with a caption. Several popular AI tools use it, such as the Stable Image of Stability AI. Human Rights Watch has found that LIAON-5B contains links to and very personal information about illegally recovered (without consent) children's stories.

>>>>>>Replacement 5182mAh L22C4PC2 for Lenovo Y7000P/R7000P

On one of the photos found in the base, we see a little girl of 2 years touching the fingers of her newborn sister. Not to mention the fact that this is a private moment that we do not want to share with just anyone, the image also contains the names of the two children and the exact address of the hospital where their mother gave birth 9 years ago. Because that is another concern noted in the report: most reactors were online years before LAION-5B existed.

The program collected them in several places: personal blogs of mothers or parents, harassment sites, even photos contained in YouTube videos, from the time when the platform was used to share things with relatives. Hye Jung Han, who found these images, says the vast majority were not accessible online via a reverse search. The fact that they were used to power AI tools is not without consequences.

>>>>>>Replacement 4325mAh A2850 for Apple iPhone 14 Plus

The reasoning is simple: an artificial intelligence is able to reproduce the photos it is given as a "meal." In other words, all the children present in the LAION-5B database potentially run the risk of a ill-intentioned person using them, for example to create deepfakes. Human Rights Watch's report recalls that in Brazil, "At least 85 girls [...]reported being harassed by their classmates, who used AI tools to create sexually explicit deepfakes of girls from photos taken on their social network profiles, and then disseminated these fake images online."

When contacted, LAION responds that it has removed the incriminated links from its database, while indicating that children and their parents are responsible for whether or not photos are posted on the Internet. Human Rights Watch calls for broader measures, with specific legislation in place. "Generative AI is still a nascent technology, and the harm already done to children is not proven. Protecting the confidentiality of children's data now will help shift the development of this technology to one that promotes, rather than violates, children's rights," concludes Hye Jung Han.

AIs were illegally trained with photos of children, according to this report