Meta AI is finally coming to the EU, but with limitations

Amid an ongoing regulatory battle with European privacy authorities, Meta announced on Thursday that its AI-powered virtual assistant, Meta AI, is finally launching in the European Union. The chatbot-like tool will be rolled out across Meta’s portfolio of social platforms, albeit with a more limited feature set compared to what it offers in its domestic U.S. market.

Separately, Meta also confirmed to TechCrunch that Meta AI will be arriving in WhatsApp in the U.K., having so far been limited to Facebook, Instagram, and Ray-Ban Meta glasses since its launch there in October.

Meta AI has been available in the U.S. since 2023, serving as an AI assistant capable of not just chatting and answering questions, but generating images, creating stylistic selfies, among other creative wonders. These features aren’t yet available in the European version.

Last month, a chat-based version of Meta AI also landed in a handful of countries across the Middle East and Africa. And now, starting this week, Meta AI will be rolling out to all 27 EU countries, plus an additional 14 European countries (and 21 overseas territories) including Iceland, Norway, Serbia, and Switzerland.

As well as being able to chat one-to-one with the assistant across Meta’s various apps, Meta AI will be made available in group chats, too, although the launch will be staggered — with the feature slated to appear first on WhatsApp, in both the EU and — as noted already — the U.K.

Privacy pushback

The EU launch of Meta AI marks the latest effort by Facebook’s parent company to spread its AI across the bloc in the face of regulatory concerns about tapping user data to train AI models.

While Meta has been training its AI on user-generated content in the U.S. for years, the tech giant (among others) has faced pushback in the EU due to the bloc’s comprehensive privacy regulations — including the General Data Protection Regulation (GDPR) — which mean it needs to have a valid legal basis to process people’s information to train AI models.

Despite these challenges, last May Meta began notifying regional users of an upcoming privacy policy change that informed them it would start using content from their comments, interactions, status updates, photos and captions for AI training. The company argued that this data processing was necessary for its AI model to reflect “the diverse languages, geography and cultural references of the people in Europe.”

However, in June, Meta was compelled to put these plans on ice following scrutiny by the Irish Data Protection Commission (DPC), Meta’s lead data protection regulator in the EU, which raised concerns about the way Meta was soliciting consent from users to process their data.

In short, Meta had implemented an onerous opt-out process (meaning users had to take action to prevent their data becoming AI training fodder) rather than a simple opt-in — with the company relying on a GDPR legal basis known as “legitimate interests,” claiming its actions were compliant with the law. However the DPC disagreed and Meta was forced to rethink its approach.

The company confirmed to TechCrunch that the version of the AI assistant it’s launching in the EU has not been trained on local users’ data — hence why it said it won’t be notifying users or otherwise seeking their consent, since it claims the technology has not been trained on their information.

“The model powering these Meta AI features wasn’t trained on first-party data from users in the EU,” Anna Dack, Meta’s innovation communications manager, EMEA, told TechCrunch.

Meta initially faced similar regulatory concerns in the U.K., which (since Brexit) sits outside the EU — but does still have a data protection regime that’s based on the GDPR.

Last summer, the U.K.’s Information Commissioner’s Office (ICO) asked Meta to pause its AI training plans over concerns about how it was helping itself to user data. But after the company adjusted its opt-out process, making it mildly less onerous, Meta went on to launch Meta AI in the U.K. — without explicit objection from the U.K. regulator, though the ICO said it would “monitor the situation”.

When asked if Meta’s AI efforts were yet trained on U.K. users’ data, a spokesperson directed TechCrunch to its announcement post back in September, where it said that it will begin training its models on user content “in the coming months,” suggesting that it’s not yet ready for the public stage.

‘Intelligent chat’

For now, Meta AI in the EU will be limited to what the company bills as an “intelligent chat function” in six European languages: English, French, Spanish, Portuguese, German, and Italian.

As it stands, the tool is basically a chatbot baked into Meta’s various apps, including WhatsApp, Instagram, Messenger, and Facebook itself. How it works is you tap a little blue circle icon to summon the assistant, and ask it any question that you might ask in a search engine, such as how to carry out a task, or to find out information on a topic.

As noted above, the feature will also be landing in group chats but this launch is being staggered — starting with WhatsApp, before expanding to Messenger and Instagram Direct Messaging “soon.”

Users will be able to call on the assistant by typing “@MetaAI”, and then asking a question — for example, where to go for dinner, or the top tourist attraction in a given city.

While Meta says in its announcement that Meta AI has “an advanced understanding of what you’re looking for,” it says this is not in reference to any kind of personalized suggestions based on the user’s data; the marketing claim is merely in the context of searching for content more easily and intuitively, it suggests.

However, the company does stress that this launch represents its “first step” in its efforts to bring more AI to Europe, and that it plans to eventually “find parity with the U.S.” over time. What this likely means is more tussles between Meta — which has been critical of Europe’s AI regulations — and EU regulators.

TechCrunch reached out to the DPC to ask for its response to the Meta AI for EU announcement.

“The DPC, as Lead Supervisory Authority for Meta, has been examining Meta AI over recent months with our colleague Supervisory Authorities across the EU/EEA and we will keep it under review as it rolls-out to users over the coming weeks,” a spokesperson told us.