Meta has defended its use of Australian users’ social media posts to train artificial intelligence without express consent, arguing that only publicly available data from its platforms is fed into the models.
The Facebook and Instagram owner also maintains that the practice is not scraping because it is first-party data, even as the company postpones AI training on European and Brazilian users’ data in response to privacy concerns.
Last week, Meta admitted to using the public posts and comments on Facebook and Instagram of its Australian users from as far back as 2007 to train its large language model, dubbed Llama.
Want to know how this story ends?
Become a subscriber for complete access to all stories on InnovationAus.com and our daily newsletter.
Do you know more? Contact James Riley via Email.
As long as the regulatory environment, the EULA and consented privacy controls allow, the most valuable asset that any SaaS product has is the data. It is first party data being used for the specific purpose of enhacing the platform. It’s a value exchange and any Meta or Microsoft property users that complain about this should consider the premise of the product and their presence there to begin with. User will go elsewhere if the platform doesn’t provide the capabilities they expect, so limiting the use of first party data in a platform for the specific purpose of enhancing it is ultra conservative and sets a precendence that not only impacts innovation, but handcaps the whole sector.