Meta wanted to train its AI with user data from Facebook and Instagram. After resistance from data protection advocates, the launch of the AI in Europe was paused for the time being.
Even though the US company behind Facebook and Instagram, Meta, has postponed the launch of its AI software in Europe for the time being, the so-called Meta-AI is not completely off the table here. Meta still believes that its approach complies with European laws and regulations. Data protection advocates see things differently.
What it's about: The company wanted to use user posts on its social networks Facebook and Instagram for its artificial intelligence (AI). However, data protection advocates criticized the fact that users could only object to this, instead of being explicitly asked for their consent. Various instructions on how to object are circulating online.
The topic remains important for users – because it also sheds new light on the issue of data protection. Answers to important questions.
Similar to other AI software – for example ChatGPT – the Meta AI creates texts and images and can answer questions from its users. To ensure that this works as well as possible, the AI is trained with data – the more, the better. And for this training, Meta also wanted to use user posts from Facebook and Instagram. For AI expert Gregor Schmalzried, this would be a competitive advantage over other AI providers simply because of the amount of usable data.
Another advantage that Meta would have from the posts from Facebook and Instagram: They are local data. “Many of the large language models have a certain bias towards the American world, because most of the data with which these models were trained comes from there,” explains Schmalzried. One example of this is ChatGPT. “Sometimes you read a sentence that feels like it was really translated word for word from English,” Schmalzried continues.
Read also
You can tell that ChatGPT is primarily an English-language model. That's different with Meta, with posts from all over the world to train the AI. The company itself emphasizes this. When Meta announced that it would postpone the launch of Meta-AI in Europe, the company argued that it would not be able to offer users in Europe a first-class experience without training with local data.
In countries such as the USA, Canada, New Zealand and Australia, however, Meta-AI is already available. It is also integrated into other Meta applications such as Whatsapp and Facebook Messenger. Users can ask Meta-AI questions there, which the AI should then answer. Unlike ChatGPT, the AI software also uses search engines such as Google and Bing.
The reason why things were different in Europe is due to data protection law. The company may have taken into account that things could be more difficult here: According to Schmalzried, European users were the only ones given the option to object to the use of their posts. But even the option to object was not sufficient for the EU, according to data protection officers. The Hamburg Data Protection Authority states: “In the case of truly private posts that were only shared in small groups, such a change of purpose would only be possible with the corresponding consent.”
In other words, Meta would have needed the explicit consent of users instead of just giving them the opportunity to object. The Federal Association of Consumer Organizations (vzbv) also criticized the fact that the objection is too complicated and that users have to justify it. For public posts, European data protection authorities are currently still examining whether Meta also needs the active consent of users here. Meta, in turn, is convinced that it has a “legitimate interest” in training the AI models and can therefore use the data accordingly.
Read also
Important for Instagram and Facebook users to know: Although Meta has stopped the introduction of its AI in Europe for the time being, it is still possible for all users to object to the use of user data. The Federal Association of Consumer Organizations explains how to do this on its website.
In general, the Hamburg Data Protection Authority advises against transmitting personal data to an AI if the provider allows the use of the data for its own purposes in its terms and conditions. This is because providers often use inputs that users make in dialogue with an AI chatbot to further train the AI. However, this can sometimes be turned off. For some chatbots, this is only possible in the paid version.
However, you can also deactivate the whole thing in the free version of ChatGPT. To do this, select the Data Controls menu item in the settings. There you can make the corresponding setting under the item “Improve the model for everyone”.
Until data protection concerns are resolved, Meta-AI is unlikely to launch in Europe. According to expert Schmalzried, this is not an isolated case: “Claude, for example, in my opinion the best chatbot in the world, was not available in Europe for a long time. The same goes for Google Gemini.”
The problem is that good AI has not yet been built in Europe, he says. And that is why Schmalzried is asking whether international companies will actually continue to accept the regulations in Europe and adapt their AI software to them with delays. Or whether they will simply no longer offer their tools in Europe because it is no longer worthwhile. Schmalzried says: “I wouldn't worry about that at the moment, but in the long term we should at least think about this scenario.”
Read also
Sources: Gregor Schmalzried is a blogger and journalist who deals with the topic of artificial intelligence, among other things. He is part of the moderator team of the ARD AI podcast. The Hamburg Commissioner for Data Protection and Information Security is the data protection authority responsible for the Meta Group in Germany.
This article first appeared on Welt.de.
Source: https://www.businessinsider.de/gruenderszene/business/meta-ki-in-europa-pausiert-was-ihr-trotzdem-wissen-solltet/
