Meta restarts training its AI systems on UK users' public Facebook and Instagram posts, having “incorporated regulatory feedback” to be “even more transparent”
Meta had already said it used public Facebook and Instagram material to train an AI assistant, while taking steps to filter private details. The UK restart follows that earlier disclosure about public-post training but adds a regulatory-feedback and transparency layer.
Meta can again use UK users' public Facebook and Instagram posts to train its AI systems, subject to the revised transparency approach it says reflects regulatory feedback.
UK users' public posts become part of Meta's AI-training input pool again; the reported change does not extend to private content.
Second-order effects
Meta's product and compliance teams must operationalize clearer disclosures around data use, making transparency a practical condition of maintaining access to social-platform training data.
The differing UK and EU outcomes increase the value of jurisdiction-specific data-governance processes for platforms training models on user-generated content.
Third-order effects
If regulators continue to permit public-content training only with stronger notice and objection mechanisms, public availability will increasingly function as a conditional permission boundary rather than an unrestricted AI-data right.
The pattern could favor large platforms able to run region-specific compliance and training workflows, while leaving model developers without comparable first-party content more dependent on licensed or open-web sources.
The trend: AI training is moving toward jurisdiction-specific governance of public user content, with transparency and regulatory engagement determining whether platforms can use it.
Unfortunately, our plans to train our AI models to understand the EU's rich cultural, social and historical contributions remain paused while EU regulators remain unable to agree how the law should be applied. As Mark and Daniel Ek, CEO of Spotify, recently warned, the EU risks …
You might scoff that this means Llama 3 won't understand the “rich cultural social and historical” contributions of Europe without access to mining public posts. But how quickly we forget the risks https://www.reddit.com/... RE: https://www.threads.net/...
Fascinating to see the mass scraping of European user activity described as an attempt to “understand the EU's rich cultural, social and historical contributions”
Large language models are architecturally incapable of <understanding> beyond anything beyond the most simplistic meaning of the word. Stop projecting capabilities onto them that they cannot and will not possess.
Meta is pleased to announce that they are done pretending, and are going to unapologetically exploit you even further, with thanks to your governments' inaction.
Unfortunately, our plans to train our AI models to understand the EU's rich cultural, social and historical contributions remain paused while EU regulators remain unable to agree how the law should be applied. As Mark and Daniel Ek, CEO of Spotify, recently warned, the EU risks …
Super ironic - in referendum I said the EU wd sabotage tech & UK wd gain from being outside, Clegg, FT chums etc laughed - now Clegg himself has to admit the Big Brexit Benefit from being outside EU's tech/regulation self-sabotage
Appalling decision. Facebook and Instagram are awash with harmful content, which will now get baked in to Meta's AI products. History repeats and replicates, seemingly with regulatory consent
Thanks to constructive exchanges with regulators, Meta's AI systems can now be trained on content publicly shared on FB and Instagram by adults in the UK and Brazil. This will enable Llama and other Meta AI platforms to better understand local languages, cultures, value systems,
The meta information that comes with a news link being shared or the snippet of content or anything that requires a request beyond Facebook to generate on their platform will be argued as a violation. 2/2
Putting aside this benefit of Brexit, press should get clarification on the word, “sharing,” as Facebook has a history of using it loosely. If they're training on any content beyond what is posted on their platform originated by users, expect significant lawsuits. 1/2 [image]
Watch out for the snake oil. “The ICO has not provided regulatory approval for the processing and it is for Meta to ensure and demonstrate ongoing compliance.” One wonders why the ICO has chosen to diverge from its EU counterparts.
We're pleased to announce that, after taking onboard regulatory feedback, we will begin training our AI at Meta models using public content shared by adults on Facebook and Instagram in the UK and Brazil.