Unlock Editor's Digest free of charge
FT editor Roula Khalaf selects her favourite stories on this weekly newsletter.
Meta warned that the EU's approach to regulating artificial intelligence posed the “risk” of cutting off the continent from access to cutting-edge services because the bloc continued to attempt to limit the ability of huge tech firms.
Rob Sherman, deputy privacy officer and vice chairman of policy for the social media group, confirmed a report that it had received a request from the EU data protection authority to voluntarily pause the training of its future AI models using data from the region.
He told the Financial Times that this could give local regulators time to “engage with the difficulty of generative AI.”
While the Facebook owner complies with the request, Sherman said such moves create a “gap within the technologies available in Europe in comparison with the remaining of the world.” He added that with future and more advanced AI releases, “it is probably going that availability in Europe could possibly be impacted.”
Sherman said: “If jurisdictions fail to manage in a way that offers us clarity about what is anticipated, then it should be harder for us to supply probably the most advanced technologies there… that may be a realistic end result that concerns us.”
These comments come at a time when big tech firms and start-ups are vying to market AI products, but are constrained by the EU's sweeping digital regulations, including the brand new Artificial Intelligence Act, which goals to manage the event of probably the most powerful models and services.
This latest request from the EU is especially as a consequence of uncertainty about whether training AI models using consumer data is permitted under the principles of the EU's General Data Protection Regulation (GDPR). These rules require firms that collect or use personal data to acquire the consent of the person concerned and to reveal the explanation for doing so.
Sherman said Meta couldn’t “adequately serve European consumers” without the flexibility to coach with European data since the AI can be less effective and unable to answer the “cultural concepts and contexts they need.”
Meta has already delayed the launch of its Meta AI assistant and a new edition of its Ray-Ban Meta smart glasses with a built-in assistant as a consequence of regulatory and privacy concerns within the EU and UK. The assistant is now available in 22 countries, including the US, Australia and Argentina, and in recent languages similar to French, Spanish and Hindi.
Apple, meanwhile, has announced that it should not launch several features under its Apple Intelligence brand as a consequence of concerns related to the EU's Digital Markets Act.
The EU said it could not comment on individual decisions made by firms, but added that it was an “attractive market with 450 million potential users that has at all times been open to any company wishing to supply services within the European single market” so long as it complied with the law.
Although Meta has stopped training future models, the corporate on Tuesday released an updated version of its AI models – Llama 3.1 – which will probably be available in Europe and worldwide.
This features a recent 405 billion parameter model, which Meta calls “the most important and strongest open model the industry has ever developed.” The next variety of parameters is usually considered more powerful.
