iiinova — where your digital self keeps evolving

Ed Newton-Rex @ednewtonrex.bsky.social · 08/10/2026

Powerful statement from mathematicians around the world. “We urge mathematicians to discontinue their work with OpenAI and to return to a vision of science that centers human understanding.” The backlash to AI is only just beginning. terrytao.wordpress.com/2026/10/07/a...

6434166

Ed Newton-Rex @ednewtonrex.bsky.social · 29/09/2026

This is exactly what copyright law is supposed to protect against: people's work being used to compete with them. The remarks were in his interview here: www.youtube.com/watch?v=v053...

youtube.com

POLITICO Pub 2026 | Day 3 at Labour Party Conference - Morning Sessions

YouTube video by POLITICO Europe

160

Ed Newton-Rex @ednewtonrex.bsky.social · 29/09/2026

Generally, ministers are intentionally vague, cushioning their language with phrases like "unless an exception applies". But the law is clear: this kind of unlicensed training is illegal. And a government minister is now on record saying so.

161

Ed Newton-Rex @ednewtonrex.bsky.social · 29/09/2026

He said this at the Politico Pub at the Labour Party conference. It is, I think, the first time that a government minister has admitted that training AI models on copyrighted work without a licence is illegal in the UK.

162

Ed Newton-Rex @ednewtonrex.bsky.social · 29/09/2026

The UK's AI minister just made a big and important admission on AI & copyright, which creatives around the country will celebrate: "It is currently illegal to develop a frontier large language model based on the transformer architecture given the copyright position that we have in this country”.

34626

Ed Newton-Rex @ednewtonrex.bsky.social · 29/09/2026

Last year, creatives across the UK fought back against the government’s unfair plans to sacrifice them at the altar of AI - and won. We should all get behind creatives in Australia as they now fight the same fight on the other side of the world. www.theguardian.com/australia-ne...

0243

Ed Newton-Rex @ednewtonrex.bsky.social · 28/09/2026

The UK government must make it clear to the Australian government that it stands with Australian and British creatives, and that legalising the theft of our work to fuel billion-dollar US tech companies is unacceptable. insidestory.org.au/will-austral...

insidestory.org.au

Will Australia become AI’s copyright-busting safe haven? • Karen Middleton

The AI giants want the federal government to okay a backdoor way of seizing creative material from around the world

0160

Ed Newton-Rex @ednewtonrex.bsky.social · 28/09/2026

If the Australian government weakens copyright law at the behest of AI companies, they would be legalising the theft of the work of the entire world's creatives on their soil.

1288

Ed Newton-Rex @ednewtonrex.bsky.social · 20/09/2026

I suspect we will soon see more lawsuits around AI companies’ use of ‘synthetic data’. Full complaint: www.musicbusinessworldwide.com/files/2026/0... /end

musicbusinessworldwide.com

081

Ed Newton-Rex @ednewtonrex.bsky.social · 20/09/2026

- expand the list of works they claim have been infringed to over 60,000 (from ~600 in their original lawsuit), which theoretically increases Suno’s maximum potential liability to around $9 billion 4/5

191

Ed Newton-Rex @ednewtonrex.bsky.social · 20/09/2026

- say this amounts to “launder[ing]” [the infringement] 3/5

181

Ed Newton-Rex @ednewtonrex.bsky.social · 20/09/2026

In the new lawsuit, the labels: - allege Suno v6 was trained on outputs of previous “tainted” models (which themselves were trained on the labels’ music) 2/5

191

Ed Newton-Rex @ednewtonrex.bsky.social · 20/09/2026

This new lawsuit from Universal & Sony against Suno is big, because it makes an argument many have made previously outside of court: that training on ‘synthetic data’ should not get you off the hook for copying people’s work for AI training. 🧵 1/5

13315

Ed Newton-Rex @ednewtonrex.bsky.social · 19/09/2026

“We trained GPT-3 on pirated stuff! No sharing that!” Astonishing admissions in these newly released documents in Authors Guild v OpenAI. www.courtlistener.com/docket/67810...

23015

Ed Newton-Rex @ednewtonrex.bsky.social · 17/09/2026

Upending creatives’ rights in the name of national security is a lazy and poorly-thought-through idea that would fail to deliver on its goals but would be sure to harm the world’s creative community. It should be fought at all costs. /end

0296

Ed Newton-Rex @ednewtonrex.bsky.social · 17/09/2026

... But they push for a broad copyright exception that gives them the right to train on *everything* because that is better for them commercially - and governments listen because they are afraid and largely clueless. 4/5

1244

Ed Newton-Rex @ednewtonrex.bsky.social · 17/09/2026

2. Most creative work is totally unnecessary to building AI for national security. As an obvious example, if you want AI for cyber security, you don’t need to train on music. The AI companies know this... 3/5

1245

Ed Newton-Rex @ednewtonrex.bsky.social · 17/09/2026

1. Permitting training on people’s work in your country does nothing for your national security. Your country will still not own the most powerful model - so what have you gained? If you think AI presents a national security concern, having the second best model does zero to protect you. 2/5

1235

Ed Newton-Rex @ednewtonrex.bsky.social · 17/09/2026

It’s clear that a number of governments are going to try to force through laws that hand creatives’ work to AI companies, on the grounds of ‘national security’. But this is an abysmally weak justification for stripping away creatives’ rights. Two core reasons why: 🧵 1/5

14321

Ed Newton-Rex @ednewtonrex.bsky.social · 09/09/2026

If that is what they’re doing, it shouldn’t be allowed IMO. AI companies should not only shut off models trained on people’s work without permission - they should not be allowed to use the outputs or data generated from those models. www.axios.com/2026/09/09/s... /end

axios.com

Suno launches new AI music models with Warner, BMG

The startup continues to face copyright lawsuits from other parts of the music industry.

2186

Ed Newton-Rex @ednewtonrex.bsky.social · 09/09/2026

They also say the new models are trained on “preference learnings from its earlier models”. The same applies - if this is user preference data on existing Suno tracks, then they have it thanks to models trained on musicians’ work without permission. 3/4

182

Ed Newton-Rex @ednewtonrex.bsky.social · 09/09/2026

It sounds to me like they have trained the new models on Suno output from previous models - output made with the old models, which were trained on musicians’ work without permission. 2/4

1103

Ed Newton-Rex @ednewtonrex.bsky.social · 09/09/2026

Suno is moving to models trained on music it has licensed. This is a big win for musicians, and likely wouldn’t have happened without the lawsuits of the last couple of years. But they say the new models were still trained on “Suno user data”. This is worrying. 🧵 1/4

22514

Ed Newton-Rex @ednewtonrex.bsky.social · 08/09/2026

AI music company Suno has been sued again - this time by the Society of Composers, Authors and Music Publishers of Canada (SOCAN). SOCAN shared examples of Suno’s output. Here is one: www.socan.com/socan-ai/ai-...

12112

Ed Newton-Rex @ednewtonrex.bsky.social · 04/09/2026

This is a full-scale assault on creators' rights. Countries should stand up to the current US administration, and refuse to legalise the theft of creators' work to benefit AI companies. www.reuters.com/business/med...

6303185

Ed Newton-Rex @ednewtonrex.bsky.social · 03/09/2026

... given that it is investing taxpayer money in billion-dollar AI companies. Sharing the full (redacted) document here so that people can take a look themselves: ed.newtonrex.com/s/Annex-A-So... /end

ed.newtonrex.com

042

Ed Newton-Rex @ednewtonrex.bsky.social · 03/09/2026

Obviously lots of caveats to this: there may be an updated business case document I haven't seen, Sov AI may have gone above their typical ticket sizes, etc. But I thought it was interesting to get more insight into the strategic thinking behind the unit... 13/n

141

Ed Newton-Rex @ednewtonrex.bsky.social · 03/09/2026

Isomorphic Labs - 0.6% Callosum - valuation unknown These do not seem like stakes big enough to achieve Sov AI's primary aim of anchoring these companies in the UK. 12/n

131

Ed Newton-Rex @ednewtonrex.bsky.social · 03/09/2026

6. Sov AI's typical ticket size is £1-9M It has invested in some big rounds (with valuations in the billions). By my calculations, if it has stuck to this ticket size, that makes its maximum stake in each investment as follows: Olix - 0.4% CuspAI - 0.5% Ineffable Intelligence - 0.2% 11/n

131

Ed Newton-Rex @ednewtonrex.bsky.social · 03/09/2026

5. There is little evidence AI firms are leaving the UK. As the document explains, "As an emerging sector, clear data showing the number of AI companies that have already left is limited." 10/n

142

Ed Newton-Rex @ednewtonrex.bsky.social · 03/09/2026

Generating a financial return and creating jobs are secondary objectives. 9/n

141

Ed Newton-Rex @ednewtonrex.bsky.social · 03/09/2026

4. Sov AI's primary aim is anchoring significant AI companies in the UK. I have seen several aims of the Sov AI Unit given in public, but this document is clear: its primary objective is to "catalyse the anchoring of at least five strategically significant AI firms in the UK over the next 10 years"

132

Ed Newton-Rex @ednewtonrex.bsky.social · 03/09/2026

... meaning the same investment gets a lower stake in the company. 7/n

141

Ed Newton-Rex @ednewtonrex.bsky.social · 03/09/2026

3. The Sovereign AI Fund was supposed to invest in Seed and Series A rounds. But its first investments Sov AI has made have been in two Seed rounds and three Series B rounds. Series B rounds are generally at higher valuations (these have been $1B+)... 6/n

142

Ed Newton-Rex @ednewtonrex.bsky.social · 03/09/2026

Sure enough, of the 5 companies Sov AI has invested in so far, 4 seem to be headquartered in London, and 1 in Cambridge. 5/n

132

Ed Newton-Rex @ednewtonrex.bsky.social · 03/09/2026

2. The UK's Sovereign AI Programme will plausibly most benefit companies in London, the South-East and the East of England. This feels obvious, but it was good to see it acknowledged, as I haven't seen Sov AI say this publicly. 4/n

152

Ed Newton-Rex @ednewtonrex.bsky.social · 03/09/2026

... I was slightly surprised to see no Net Zero or Climate Change Mitigation Assessment conducted. I don't know whether these would usually be carried out, but I thought it was striking that they weren't in this case. 3/n

143

Ed Newton-Rex @ednewtonrex.bsky.social · 03/09/2026

The Sovereign AI Fund has said it is handing out "up to one million GPU hours per company", and lists one of its key risks as "dependencies on national compute/AI infrastructure". Given that the environmental impacts of AI data centres are such a hot topic... 2/n

152

Ed Newton-Rex @ednewtonrex.bsky.social · 03/09/2026

I got hold of the Sovereign AI Unit's 'Outline Business Case', the document used to get the UK government's sign-off on initial funding, via an FOI request. A *lot* is redacted, but some things jumped out: 1. No environmental assessments 1/n

396

Ed Newton-Rex @ednewtonrex.bsky.social · 27/08/2026

I tweeted about this over a year ago, and the Hugging Face CEO suggested I file a report on their site. I did so. The report seems to have been ignored, and the dataset continues to be downloaded thousands of times every month. /end

0349

Ed Newton-Rex @ednewtonrex.bsky.social · 27/08/2026

Its employees added a dataset of 7,000 books to the platform that were taken by others from Smashwords, an ebook site. The briefest of glances shows many of them contain clear copyright notices. 2/3

12812

Ed Newton-Rex @ednewtonrex.bsky.social · 27/08/2026

Hugging Face, which is being acquired by Nvidia for $13 billion, hosts and distributes large datasets of copyrighted books without permission. 🧵 1/3

36742

Ed Newton-Rex @ednewtonrex.bsky.social · 26/08/2026

Surprised AISI didn’t publish it tbh

120

Ed Newton-Rex @ednewtonrex.bsky.social · 26/08/2026

2. Make it clear that AISI is liable for harm caused by the models it is testing. This should improve monitoring and other precautions significantly, and will make the public more safe. 6/7

183

Ed Newton-Rex @ednewtonrex.bsky.social · 26/08/2026

In my mind, government should do two things: 1. Make it compulsory for AI companies to submit their models for testing. This would remove AISI's reliance on them and the perverse incentives that brings. 5/7

141

Ed Newton-Rex @ednewtonrex.bsky.social · 26/08/2026

"Rather than providing a robust check on the private sector, the government agency becomes captive to the companies it is supposed to monitor because it is dependent on their good will to continue to function at all." 4/7

132

Ed Newton-Rex @ednewtonrex.bsky.social · 26/08/2026

It also points out it could be "safety washing" AI companies, making them seem more safety-conscious than they really are: the companies provide their models for testing voluntarily, so AISI is disincentivised from doing anything that would annoy them. 3/7

112