Meta, the parent company of Facebook and Instagram, has announced its plans to use public content shared by European users on its platforms to train its AI tools, which are known collectively as “AI at Meta.” This move is part of the company’s plan to roll out these tools in Europe, as they are currently available only in selected countries. Meta states that the training from public content is necessary to ensure the AI models accurately understand and respond to regional languages, cultural nuances, and trending topics on social media.
However, the plan has come up against EU data protection laws, known for their strict approach to personal data protection. The use of public content from European users to train AI models has sparked debates regarding possible breaches of GDPR privacy laws. Despite these discussions, Meta plans to enforce updated terms of service from 26 June 2024 in Europe.
Privacy advocacy group, None Of Your Business (NOYB), has instigated legal challenges against Meta by filing complaints in 11 European countries in a bid to halt the planned changes. Max Schrems, the founder of NOYB, criticizes that Meta’s plan, essentially, enables it to use ‘any data from any source for any purpose and make it available to anyone in the world.’ Schrems argues that this breaches GDPR compliance as “AI technology” is an extremely broad term with no precise legal boundary, raising concerns that the user data could be used for anything from chatbots to personalized advertising or even weaponized drones.
While Meta has proposed an opt-out form for users, Schrems counters this, advocating instead for an opt-in approach and condemning Meta’s opt-out form as “hidden and misleading”. If Meta proceeds with its plan, any content from dormant or old European Facebook and Instagram accounts will also be added to the AI training dataset unless users explicitly opt-out.
Meta insists it does not use people’s personal messages to family and friends to train its AI systems, and argues that other companies such as OpenAI and Google already train their models with public data. Therefore, it believes, European users should have no issues with its own similar approach. However, this argument has yet to assuage the concerns of privacy advocates and those worried about potential GDPR infringements.
The ongoing controversy underscores the complex interplay between advancing AI technology and respecting user privacy, especially within regions with strict data regulations like Europe. The decisions reached in this case could have far-reaching implications for the future use of public data in AI training.
