Meta plans to start using data collected from its users in the European Union to teach their AI systems, company announced todayStarting from this week, the technical giant will begin to notify Europeans by e -mail and its family of applications about the fact, with the message that will be included in order to include the explanation of the type of data that he plans to use as part of the training. In addition, the notification will refer to the form that users can complete to abandon the process. “We made this form of objections easily to find, read and use, and we observe all the forms of objections that we have already received, as well as recently presented,” says meta.
The company notes that it will use only the data that he collects from public messages, and META AI interaction for educational purposes. He will not use private messages in his training sets, as well as some kind of interaction, public or other, made by users under the age of 18 years. As far as the company wants to start using EU data now, it claims that the information will allow it to accurately configure its future models in order to better serve Europeans.
“We believe that we are responsible for the creation of AI, which is not only available to Europeans, but also actually built for them. That is why it is so important that our generative models of AI be trained in various data so that they can understand the incredible and diverse nuances and difficulties that make up European communities, ”the meta says.
“This means everything, from dialects and conversations, to hyper-local knowledge and various methods that different countries use humor and sarcasm on our products. This is especially important, since AI models become more advanced with multimodal functionality, which covers text, voice, video and images. ”
Meta Notes other artificial art companies, including Google and Openai, have similar data from European users to accurately configure their own systems. Today's announcement follows the initial release of Meta's New Lama 4 Models. After early excitement, the company was accused of Lmarena gamesThe website on which people compare the results of various AI models to evaluate them. Researchers noticed that META provided the experimental version of Llama 4 on the site “optimized for conversation”.