Meta resumes plans to use UK public posts for AI training

Meta has announced that it will resume its efforts to train artificial intelligence systems using public posts from its UK users on Facebook and Instagram. This move marks a significant step forward for the company, which had previously paused these plans due to regulatory concerns in the UK and the European Union.

The decision to restart training its AI with UK public data comes three months after Meta halted the initiative in response to scrutiny from the UK’s Information Commissioner’s Office (ICO) and the Irish Data Protection Commission (DPC). The ICO had expressed concerns about how Meta planned to use UK user data for training its generative AI algorithms and about the consent process. The DPC, acting as Meta’s lead regulator in the EU, also raised objections following feedback from various data protection authorities across Europe.

While Meta has been using user-generated content in the US for AI training for some time, the company’s efforts in Europe have been impeded by stringent privacy regulations. In May, Meta notified EU users of a planned update to its privacy policy, which would allow it to use content from comments, status updates, photos, and other interactions for AI training. This move triggered complaints from the privacy activist organization NOYB (None of Your Business), arguing that Meta’s approach contravened GDPR regarding consent.

Meta has defended its actions by invoking the GDPR provision of “legitimate interest,” though this defence has been previously challenged by the Court of Justice of the European Union (CJEU) in the context of targeted advertising. The decision to restart AI training in the UK, rather than the EU, reflects the UK’s continued alignment with GDPR principles, even post-Brexit.

Meta asserts that it has adjusted its approach based on regulatory feedback to enhance transparency. Starting next week, UK users will receive in-app notifications about how their public content will be used. Meta promises that this will allow its generative AI models to better reflect British culture and idioms, and will enhance the technological capabilities available to UK businesses and institutions.

Objections to Meta’s approach remain, particularly concerning the process for users to opt out of data usage. The company previously required users to navigate a complex objection form rather than providing a straightforward opt-in or opt-out option. Meta has stated that it has simplified the objection form this time, though specifics have not yet been provided.