Meta's AI Training Gets the Go-Ahead in the UK
After halting its AI system development in July due to concerns from UK authorities, Meta has now received approval to resume the use of public user posts in its AI training efforts. This follows negotiations with British regulators, allowing Meta to utilise public content from Facebook and Instagram.
Meta’s Statement
According to Meta: “We will begin training AI at Meta using public content shared by adults on Facebook and Instagram in the UK over the coming months. This means that our generative AI models will reflect British culture, history, and idiom, and that UK companies and institutions will be able to utilise the latest technology.”
This statement positions Meta’s initiative as a positive step for cultural representation, although at its core, this process is about using human interaction data to train AI models. Meta, like other AI developers, requires such data to teach its models how people communicate and to improve the accuracy and context of their responses.
AI Training: More Than Culture
While Meta suggests that its AI will reflect British culture, the underlying goal is to understand and adapt to the nuances of language and communication. By framing it in this way, Meta is seeking to mitigate concerns about data usage in AI development. The reality is that this data collection helps AI systems better replicate human interactions, not necessarily cultural specificity.
Legal Framework and Data Privacy
Meta’s approval to use public posts in the UK is based on the legal provision of "legitimate interests," which ensures compliance with UK law. It is important to note that Meta is not using private messages or data from users under the age of 18 in this process. Public posts, comments, photos, and captions from adult accounts will be used to improve the company's generative AI models.
Meta has been keen to clarify this point: “We do not use people’s private messages with friends and family to train AI at Meta, and we do not use information from accounts of people in the UK under the age of 18. We’ll use public information – such as public posts and comments, or public photos and captions – from adult users on Instagram and Facebook to improve generative AI models for our AI at Meta features and experiences, including for people in the UK.”
Global Implications
Meta paused AI training in both the UK and Brazil earlier this year due to regulatory concerns. However, British authorities have now permitted the use of public data, and Brazilian regulators have also followed suit. This marks a significant step forward for Meta’s AI efforts. Nonetheless, Meta continues to face challenges in Europe, as the EU deliberates over restrictions related to the use of user data in AI training.
EU Regulations and Meta’s Response
In June, Meta was required to introduce an opt-out option for EU users who did not want their posts included in AI training. This was due to the EU’s “Right to Object” under its Digital Services Act (DSA). Meta’s leadership has expressed frustration with the stringent regulatory environment in Europe. Nick Clegg, Meta’s President of Global Affairs, stated: “Given its sheer size, the European Union should do more to try and catch up with the adoption and development of new technologies in the U.S., and not confuse taking a lead on regulation with taking a lead on the technology.”
Meta is pushing for more flexibility in AI development, arguing that data availability is crucial to building advanced tools. However, this must be balanced with users’ rights to control how their personal content is used.
User Privacy Concerns
Despite Meta’s assurances that private messages are not used for AI training, the use of public posts still raises concerns. For instance, a user publicly posting about a personal matter, such as a family funeral, might not want that information to contribute to AI training. Although the chances of this data appearing in AI-generated content are minimal, users may feel uncomfortable with its inclusion.
Tech companies have been criticised for their early approach to AI model development, which involved scraping data from various platforms with little regard for user consent. Much like the rapid growth of social media, AI development has been focused on speed and market dominance, often at the expense of considering potential harm.
A More Cautious Approach
Given these issues, it is sensible to adopt a more cautious approach to AI training. Regulators and companies alike should fully understand the implications before approving widespread data use. If users do not wish to have their public posts included in AI training, they are advised to switch their profiles to private.
