Meta has cancelled the launch of Meta AI in Europe as a consequence of objections to the plan to coach its models using users' social media data.
Earlier this month, Meta told European users of its social media platforms that it intended to make use of their publicly released data to coach its models. The proposed changes to its terms of service drew strong criticism from GDPR regulators and privacy groups.
Meta took an identical approach to OpenAI and founded its company in Ireland to be protected by the Irish Data Protection Commission (DPC) when complying with GDPR data protection laws.
But now even Ireland's data protection authority, which has been accused of helping tech firms circumvent GDPR, has said Meta's plans may go too far.
Meta's plans to make use of Facebook and Instagram posts from EU users would have come into effect on June 26, 2024, but have now been cancelled.
In an announcement, Meta said: “We are upset by the request from the Irish Data Protection Commission (DPC), our lead regulator, on behalf of European data protection authorities, to delay the training of our large language models (LLMs) on public content shared by adults on Facebook and Instagram… This is a step backwards for European innovation and competition in AI development and can cause further delays in bringing the advantages of AI to the people of Europe.”
Meta stresses that other firms operating within the EU are already implementing its proposals and that Meta is “more transparent than a lot of our industry colleagues”.
Collective punishment?
Without EU user data, Meta says, “we’d only find a way to supply people a second-class experience. This means we cannot launch Meta AI in Europe at the moment.”
The data protection organization NOYB, which has filed eleven complaints with various data protection authorities about Meta's plans, welcomed the most recent developments with caution.
(Provisional) Noyb victory: @Meta stops using EU/EEA data for “meta-AI” after 11 complaints from noyb and increasing criticism from other EU/EEA data protection authorities! 😀🥂
More: https://t.co/bTpIAhBVkI https://t.co/czThdIcNca
– noyb (@NOYBeu) June 14, 2024
NOYB Chairman Max Schrems said: “We welcome this development but will monitor it closely. So far there is no such thing as a official change to the Meta Privacy Policy that might make this obligation legally binding. The cases we’ve filed are ongoing and should be decided.”
A straightforward solution that might meet GDPR requirements can be for Meta to require users to present explicit consent via an opt-in process quite than having them opt out.
Schrems said: “Meta's press release reads a bit like 'collective punishment'. If one European insists on his rights, all the continent won’t get our shiny latest products. But Meta has every opportunity to introduce AI on the idea of valid consent – it just chooses to not accomplish that.”
This doesn’t change the undeniable fact that Meta continues to make use of data from users outside the EU without an opt-out choice to train its models, however it illustrates the facility of regulation and community resistance.
