Meta training AI on social media posts? Only 7% in Europe think it’s OK – Source: go.theregister.com

Updated Meta’s enthusiasm for training its AI on user data is not shared by the users themselves – at least for some Europeans – according a study commissioned by Facebook legal nemesis Max Schrems and his privacy advocacy group Noyb.

Noyb (None Of Your Business) got Gallup to do a poll of 1,000 Facebook and Instagram users in Germany, and found that just 7 percent wanted Meta to train its AI models on their data. Perhaps more damningly, 27 percent of the users participating in the survey didn’t even know that Meta was using their data in this way.

“Meta probably knows that no one wants to provide the data from their social media accounts just so that Meta gets a competitive advantage over other AI companies that do not have access to such data,” said Schrems, whose decade plus crusade against the company has been responsible for the collapse of two US-EU data-sharing agreements, along with significant changes to Meta’s data-collection practices in Europe.

“Instead of asking for consent and get[ting] ‘no’ as an answer, they just decided that their right to profits overrides the privacy rights of at least 274 million EU users.”

To comply with the EU’s General Data Protection Regulation and its British version, which still exists as the UK GDPR post-Brexit, Meta needs a legal basis for training its AI on users’ public posts and comments. As it would rather avoid asking for opt-in consent, its only other option has been to claim that it has “legitimate interests” for the processing.

Noyb’s complaints about this led Meta to suspend its AI training in Europe last year, but data protection authorities in the EU and the UK subsequently approved Meta’s use of the legitimate-interests basis. In May, as Meta prepared to resume EU training, a German regional court also rejected an injunction application by consumer advocates, saying Meta’s GDPR justification is kosher.

• Irish privacy watchdog OKs Meta to train AI on EU folks’ posts
• Max Schrems launches privacy NGO, wins €60k within first 24 hours
• Zuck tries to justify AI splurge with talk of ‘superintelligence’ for all
• White House bans ‘woke’ AI, but LLMs don’t know the truth

“Noyb’s misguided claims are wrong on the facts and the law,” a Meta spokesperson said in an emailed statement to The Register. “Last year we sent more than 2 billion in-app notifications and emails to people in Europe to explain what we’re doing and their right to object. We also notified everyone using Facebook and Instagram in the EU again this year.”

However, Noyb claims that fewer than half of Meta’s users saw those notifications and emails, with only 21 percent of youngsters in particular remembering them. The organization’s take is that this undermines Meta’s legitimate-interests justification, which partly relies on not confounding the “reasonable expectations” of users as to what will happen to their data.

Noyb is contemplating a potential class action against Meta that could cost the company billions. Meanwhile, German privacy officials are predicting that Meta’s AI practices will end up being adjudicated at the EU’s highest court. In that context, Schrems’s latest data nugget could end up serving as useful evidence.

Neither the Irish nor UK data protection authorities had provided comment on Noyb’s survey at the time of publication. ®

Updated to add:

The UK’s ICO told us: “Where organizations are relying on the legitimate interests lawful basis to process their users’ data for AI training purposes, they must provide people with clear information about these activities and give them a simple route to opt out of the processing.”