UK government signals it will not force tech firms to disclose how they train AI |
Peer accuses ministers of lying as government dismisses Lords’ request to make firms reveal what materials they use
https://www.theguardian.com/law/2025/jun/06/uk-government-signals-it-will-not-force-tech-firms-to-disclose-how-they-train-ai
Posted by MetaKnowing
11 Comments
There is £££££ loads of reasons ££££ why ministers might have changed ££££ their minds.
Piracy is bad… unless wealthy people do it then it’s completely fine.
Yet another reason why the British public is disillusioned with its country when they’re yet again putting money in front of the issues the 99% face
I see no reason why anyone should pay for subscription services anymore then. Fuck it.
Hopefully they’ll stilll tell us when it becomes self-aware.
It makes some sense if you understand LLM training relies on a curated dataset from the internet. As the ‘net fills with AI generated data, it will get more difficult to produce such a dataset. The dataset is the asset, not the research (one of the reasons papers are published openly) nor the GPUs (since nearly anyone can buy a GPU). The datasets are extremely valuable and will be worth vastly more in the future so it’s not surprising they want to keep as much about them secret as possible.
Gov have effectively now said piracy is legal. Man the sails, bois!
I work in digital sustainability and AI is a major cause for concern in the tech industry regarding the amount of pollutants and emissions it is causing. It’s already a well established fact backed by years of scientific research that ICT emissions from Internet related activity exceed the entire aviation industry (its upward of **8%** of all global emissions). AI being one of the most significant causes of emissions via energy use (often from non-green sources), water use (for cooling on scales you can barely imagine, even to the point of depriving humans of their supply), and material waste (both in the production and use of equipment and buildings for data centers) as one of the most high-intensity drivers of server uses atm.
It’s a clusterfuck of a situation that is making our efforts to battle climate change ever more futile, mostly for the sake of stuff like generative slop that is single use, often inaccurate or misleading, and so far of limited benefit (based on the scientific literature which currently shows no tangible performance or profit yields beyond stock market wishes and prayers and biased self-reporting placebos).
I was told not to download cars when I was growing up, which is impossible, but apparently since the advent of AI corporations can in fact download everything, plant it into a program and reproduce it infinitely forever. Ta.
I think eventually someone will challenge this in the courts and win and then there will be trouble.
At the very least you need to be able to know if your work was used for training or not.
Pirate what you want – just run a local llama AI instance and say you are ‘training it’.
There you go – instant defence.