Meta's 'pay or consent' model likely to be on borrowed time in EU
Is this why release of newest LLM is confined to the US?
Meta has been given until 1st September to respond to consumer protection concerns in the EU, and has confined its newest Llama model launch to the US.
The Consumer Protection Cooperation (CPC) Network, which enforces EU consuumer protection laws, is concerned about what has been termed the ‘pay or consent' model.
The licensing model changed late last year, and regional users of Facebook and Instagram are now required to make a binary choice. The choice is between consenting to being tracked and profiled for digital ad targeting if they want to continue using the platform without charge, or paying Meta a monthly subscription for ad free services.
The "pay or consent" model has already placed under considerable scrutiny by the EU. The EU Commission suspects that it contravenes the Digital Markets Act, and also possibly the Digital Services Act. The model is also being investigated by the regulators responsible for upholding the GDPR across the bloc.
Now Meta has been set a deadline of 1st September to engage with the CPC Network, which has been investigating the model since complaints began pouring in last November. The network also expects Meta to propose some solutions.
Almost 20 consumer organisations filed complaints with the CPC Network last November, and a further eight consumer group also filed complaints about what they considered to be breaches of the GDPR.
"Meta has until September 1, 2024 to reply to the letter of the CPC network and the Commission and to propose solutions. If Meta does not take the necessary steps to solve the concerns raised, CPC authorities can decide to take enforcement measures, including sanctions," the Commission said in a statement earlier this week.
The statement indicates that the consumer protection watchdogs believe that several elements of Meta's consent mechanism could constitute "misleading or aggressive practices," in contrast with the EU's standard of consumers being provided upfront "with true, clear and sufficient information."
The European Data Protection Board ruled earlier this year that the model did not allow users to exercise 'freely given consent' and the Commission statement builds on this ruling.
Firstly, the use of the word ‘free' is being contested. The CPC suspects that it is misleading for consumers because if they choose not to pay for the ad-free version their personal data will be used to create revenue for Meta via showing them personalised ads.
They also take issue with the claim that paying users will not see any ads because they are still likely to when engaging with the content of other users.
Another bone of contention is Meta's using woolly language to downplay what customers are actually consenting to, using terms such as "your info" rather than "your personal data," and the confusion created by having click through multiple screens and links to access the relevant parts of the Terms of Service and Privacy Policy.
The CPC also says the practice of essentially locking users out of their accounts which they were used to being free, until they make their choice, is not allowing them the opportunity to make a properly informed decision.
Responding to the CPC action, Meta spokesman Matthew Pollard emailed a brief statement and referenced a ruling by the EU's top court from last summer which concerns a separate competition challenge.
He said: "Subscriptions as an alternative to advertising are a well-established business model across many industries. Subscription for no ads follows the direction of the highest court in Europe and we are confident it complies with European regulation."
Llama 3.1 405B is go - and open source
It is partly the concerns about Meta's compliance with multiple pieces of EU legislation – and Meta's reciprocal concerns about the blocs more stringent framework of AI and data regulation – that has led to a US only launch of Meta's newest LLM.
Llama 3.1 405B is, according to a blog post from Meta, Competitive with OpenAI and Anthropic across a range of tasks. Ironically, and perhaps not for long, the model is available is free of charge.
"Developers can fully customise the models for their needs and applications, train on new datasets, and conduct additional fine-tuning," Meta said. "This enables the broader developer community and the world to more fully realise the power of generative AI. Developers can fully customise for their applications and run in any environment … all without sharing data with Meta."
The reaction to Meta's open-source LLM is likely to be mixed. There are certainly potential benefits, and the power of GenAI being more widely accessible rather than concentrated in the hands of a very limited number of tech giants sounds appealing. There are some good arguments in favour, and Mark Zuckerberg sets them out in a separate blog post.
However, he also concedes, "bad actors may be able to use the intelligence of AI models to fabricate entirely new harms". He goes onto argue: "I think it will be better to live in a world where AI is widely deployed so that larger actors can check the power of smaller bad actors."
Right now, access to the model is limited to the US, but that isn't likely to remain the case for long.