Hannah Murphy and Cristina Criddle are correct to categorise Meta pulling the plug on its Llama artificial intelligence model in the EU as the latest leg of Big Tech’s “race” to “commercialise AI products” (Report, July 23). However, this dash to the finish line has exposed the unsustainable business principles by which many of these models are being trained.
The EU’s stance on copyright law is well established, and its AI Act is clear. All organisations providing or deploying general-purpose AI models within the EU must document and publicly disclose a summary of the content used to train their models. Additionally, providers of these models must put in place a policy to ensure they comply with EU copyright law.
Regulation is not about putting the brakes on AI innovation; it’s about creating guardrails to ensure sustainable development. In light of this, Meta has a safe way to proceed and it is to ensure that its AI models are trained with data from individuals who have given explicit consent.
Meta has consistently emphasised its commitment to publishing models under free and open source licences, a dedication reaffirmed in its recent “open source letter”. I argue that, if Meta cannot train its models in a way that doesn’t spark “uncertainty over whether training AI models on consumer data was permissible within the EU’s General Data Protection Regulation (GDPR) rules” then Meta’s refusal to adapt its practices to align with these regulations demonstrates a disregard for the legal and ethical standards necessary to protect consumer data. What precedent does that set? Apple, Anthropic, Nvidia and other Big Tech companies have found themselves in a similar situation recently, having used YouTube subtitles to train AI models.
Thus Meta and other Big Tech companies currently stand at an important crossroads: Meta can choose to be a global leader in open and sustainable AI innovation or it can set a dangerous precedent for the industry that is difficult to undo. Ultimately, Meta must recognise that the future of AI hinges on sustainably sourced data. If not, let this serve as a wake-up call to start sourcing responsibly.
Meeri Haataja
Chief Executive and Founder, Saidot, Helsinki, Finland