Meta to start training artificial intelligence models on the general content of the European Union

Dead Declare On Monday, artificial intelligence models will be trained in general content, such as posts and comments on Facebook and Instagram, in the European Union after previously Her plans stop to do so In response to regulatory pressure due to data privacy concerns. She said that the company will start training AI on the content of users in the European Union this week. User reactions will also be used with Meta AI to train their models.

This announcement comes after a limited copy of Meta AI was launched in the European Union Last month, after appearing for the first time in the United States and other global markets.

While it was dead Artificial intelligence training On the content created by the user in the United States for years, it has faced resistance in the European Union due to the strict laws of the bloc, especially the General Data Protection Regulation (GDPR), which requires a clear legal basis to process personal data to train artificial intelligence models.

Mita said Again in June 2024 It will stop temporarily to start training AI systems using user data in the European Union and the United Kingdom after its reaction from the Irish Data Protection Committee (DPC). DPC regulates the definition of the European Union and was behaving on behalf of many of the block protection authorities. in September 2024Meta said she was restarting efforts to train artificial intelligence systems using public publications from the UK user base.

Fast forward to this day; Meta announced that it will do this with public posts from the European Union user base as well.

“Last year, we postponed the training of our large language models using general content while the organizers explained the legal requirements,” Mita said in the blog post. “We welcome the opinion of EDPB in December, which confirmed that our original approach has fulfilled our legal obligations. Since then, we have been constructed with the IDPC and we are looking forward to continuing to provide the full benefits of male individuals in Europe.”

Starting this week, European Union users will start receiving notifications in the application and email to explain that Meta will start using public data and interactions with Meta AI to train their models. These notifications will include a link to a model that allows users to cancel the subscription of their data used. Meta says she will honor all the objection models that she already received, as well as the newly presented models.

Meta notes that he does not use private messages, nor general data from users under the age of 18 in the European Union, to train their models.

“We believe that we have the responsibility to build Amnesty International not only for Europeans, but it has already been built,” Mita says. “For this reason, it is extremely important that our artificial intelligence models are trained on a variety of data so that they can understand the nuances and amazing and varied complications that make up European societies. This means everything from dialects and colloquial, to extreme local knowledge and the different ways that the humor and ridicule of our products.”

Meta says it follows the example that companies such as Google and Openai have already used data from European users to train artificial intelligence models.

Meanwhile, DPC is not fully moved from checking how adult language models are training in artificial intelligence services. Last week, the organizer Declare He was investigating the Xai Grok training.