Key Points:
- Meta will resume AI training in the UK using public content from Instagram and Facebook.
- Private messages and data from users under 18 will be excluded from the training process.
- Meta promises to simplify the opt-out process and notify UK users about data usage and opt-out options.
Meta is set to restart training its AI models using public content from Facebook and Instagram in the UK, following a regulatory pause. The training will exclude private messages and any data from users under 18, ensuring compliance with local regulations.
This decision is part of Meta’s efforts to ensure its AI models accurately reflect British history, language, and culture. Public posts like photos, captions, and comments will be used in the training process, but sensitive data remains off-limits.
In response to feedback from UK regulators, Meta claims it has made the process more transparent. The company stated in a blog post that it has worked positively with the Information Commissioner’s Office (ICO) to refine its approach. Although the ICO has not officially approved the plan, it will continue to monitor the process closely.
Starting next week, UK adults on Facebook and Instagram will be notified about how their data is used for AI training, with an opt-out option made more accessible. Previously, the opt-out process was seen as overly complicated, requiring users to follow multiple steps. Meta has now streamlined this form, integrating feedback from the ICO, and users who have already opted out will not be notified again.
Meta’s Global AI Training Plans
In June, Meta paused its AI training plans in the EU and US due to regulatory concerns raised by the ICO and the Irish Data Protection Commission. Meta criticized the EU’s restrictions, stating they hinder innovation, though it’s unclear when training will resume in these regions.
Globally, Meta continues to develop its AI models with a focus on global diversity and plans to expand AI training to include more countries and languages by the end of the year.
Recently, Meta admitted to using public data from all adult users in Australia for AI training without offering an opt-out option. While private data and content from users under 18 were excluded, it remains uncertain if data from users who turned 18 during that period was included.
Australia is also considering updates to its Privacy Act and mandatory AI guidelines, which could impact Meta’s future AI practices in the country.