Meta has postponed its plans to train AI models using public content shared by Facebook and Instagram users in the EU, following a request by the Irish Data Protection Commission. The company mentioned that it is in consultation with the IDPC and has incorporated the authority’s feedback.
This comes after the advocacy group None Of Your Business (NOYB) filed 11 complaints against it for changing its privacy policy to allow for training of its AI models on public user data. NOYB highlighted that users are not given any information about the future AI their data is being used to train, which is against the EU’s General Data Protection Rules (GDPR) requirements.
Meta has argued that it has a ‘legitimate interest’ in processing first and third-party data to build its AI services as a legal basis for processing personal data under the GDPR. However, NOYB points out that when Meta tried to make a similar case of legitimate interest in the context of using all personal data for advertising it was rejected by the European Court of Justice.
What is Meta’s response to the complaints against it?
The company says that to properly serve its customers in the EU the models that power AI at Meta need to be trained on information that reflects the languages and culture of the region. “We believe that Europeans will be ill-served by AI models that are not informed by Europe’s rich cultural, social and historical contributions,” Meta argues.
The company argues that other companies, such as Google and OpenAI, also rely on European user data to train their models, implying that this practice is common in the industry. “Our approach is more transparent and offers easier controls than many of our industry counterparts already training their models on similar publicly available information,” it argues. It mentioned that its opt-out form (which users can fill out to prevent their data from being used for training AI) is easier to find, read, and use than its industry counterparts. Responding to the use of ‘legitimate interest’ as the grounds for data processing, the company added that other industry players have also used the same.
The company also clarified the data used for training purposes will be information shared publicly on Meta platforms. While private messages will not be used as training data, the company anticipates using other content, such as interactions with AI features or chats with a business using AI at Meta AI for training purposes. It mentions that the models trained on this data are not designed to identify individuals, but rather to identify patterns “like understanding colloquial phrases or local references.”
Also read: