Meta AI is temporarily stopped in the EU
Meta Platforms, the parent company of Facebook, Instagram and WhatsApp, has just announced that it will not launch Meta AI in the European Union until further notice. The reason is – as is so often the case with Meta – concerns from data protection advocates.
Meta has agreed with the Irish Data Protection Authority (DPC), which is responsible for Meta’s European headquarters in Dublin, not to train its AI models with data from European users who have published it on Facebook or Instagram for the time being.
AI feature makes Apple more valuable than ever, bigger than Nvidia
“The DPC welcomes the decision by Meta to pause its plans to train its large language model using public content shared by adults on Facebook and Instagram across the EU/EEA. This decision followed intensive engagement between the DPC and Meta. The DPC, in co-operation with its fellow EU data protection authorities, will continue to engage with Meta on this issue,” Ireland said.
Meta AI is a standalone chatbot in the style of ChatGPT, which has the unique selling point of being able to create images based on prompts in real-time. The AI models are also integrated in other forms, for example in searches on Instagram or WhatsApp, but currently only for users in markets outside the EU. It will stay that way.
How likely is AI to wipe out humanity? 70%, according to ex-OpenAI employee
“Step backward for European innovation”
The DPC welcomes Meta’s decision to halt its plans to train its large language model on public content shared by adults on Facebook and Instagram in the EU/EEA. This decision follows intensive discussions between the DPC and Meta. The DPC, in cooperation with the other EU data protection authorities, will continue to engage with Meta on this matter.
“We are disappointed by the request from the Irish Data Protection Commission (DPC), our lead supervisory authority, on behalf of the European data protection authorities, to delay the training of our large language models (LLMs) with public content shared by adults on Facebook and Instagram – especially as we have taken on board the feedback from the supervisory authorities and the European data protection authorities have been informed since March. This is a step backwards for European innovation and competition in AI development and will lead to further delays in delivering the benefits of AI to the people of Europe,” said a statement from Stefano Fratta, Global Engagement Director and Meta’s Privacy Policy Officer.
However, they remain very confident that Meta’s approach will comply with European laws and regulations – so far no changes have been made to data protection regulations. “We are committed to making Meta AI and the models underlying it accessible to more people around the world, including in Europe. But without incorporating local information, we would only be able to offer people a second-class experience. This means that we cannot currently launch Meta AI in Europe,” they continue.
AI models should recognize patterns and not individuals
Meta apparently does not want to deviate from the practice of using user data for AI purposes. “We are following the example of others, including Google and OpenAI, who have both already used data from Europeans to train AI. Our approach is more transparent and offers simpler controls than many of our industry colleagues who already train their models on similar publicly available information,” Fratta says. “The models are built using information from people to recognize patterns, such as understanding colloquial expressions or local references, and not to identify a specific person or their information.” This would comply with European laws and regulations.
It had previously been communicated that publicly accessible content from Instagram and Facebook users (not private messages) would be used to train the LLMs – and there was also an option to object to the use of the data for these AI training purposes. Finally, the data protection NGO noyb of Viennese Max Schrems, together with other organizations, filed 11 complaints against Meta AI with various data protection authorities in Europe.
“We welcome this development, but will monitor it closely. So far, there has been no official change to the meta data protection regulations that would make this obligation legally binding. The complaints we have submitted are not yet closed and must now be decided,” said Max Schrems, chairman of noyb, in a statement. He sees the temporary end of meta AI in the EU as a success of the pressure that had been built up by the complaints.
