The UK AI Copyright Reversal: What Happened and Why It Matters
Gary WhittakerThe UK Tried to Let AI Train on Copyrighted Work. Here’s Why It Hit a Wall.
A clear deep dive on the UK’s AI copyright reversal, the September 2026 U.S. government fair-use position, Canada’s SOCAN litigation against Suno, and why copyright enforcement still matters.
JR Insights • AI Copyright Deep Dive • Updated September 3, 2026
The UK Tried to Let AI Train on Copyrighted Work.
Here’s Why It Hit a Wall.
This was never just a fight about AI hype. It was a fight over something much more concrete: who controls copyrighted work, who gets paid when it is used, and who carries the burden when that system is challenged.
September 2026 Update
The U.S. government has now taken a position. Canada has moved into litigation.
On September 2, the U.S. government filed a brief supporting OpenAI’s fair-use position in its copyright dispute with The New York Times. The Commerce Secretary also urged G20 countries to develop frameworks that allow AI training on copyrighted works while balancing creator protections. These are government legal and policy positions—not a court ruling, not a new statute, and not proof that every form of AI training is fair use.
Canada has also moved beyond consultation alone. On September 2, SOCAN filed a Federal Court action against Suno alleging infringement of performing rights in musical works. Canada still has no AI-specific copyright statute, but existing Canadian law is now being tested directly against an AI music generator. Read the JR Suno lawsuit tracker for the Canadian case.
Learn More
Want a clearer path through AI, copyright, and creator tools?
Explore free guides on how AI can help you, then join the newsletter to stay connected as this space keeps changing.
The short version
UK
The government backed away from a plan that would have let AI companies use copyrighted work by default unless creators opted out.
US
Courts still decide the cases, but the federal government has now formally backed a fair-use position for AI training in the OpenAI–New York Times litigation.
Canada
Canada still has no AI-specific copyright rule, but SOCAN is now testing existing Canadian law against Suno in Federal Court.
What happened in the UK this week
The UK government stepped back from a controversial idea: letting AI companies use copyrighted material by default for training, unless creators took action to opt out.
That mattered because it would have flipped the normal logic of copyright on its head. Instead of asking first and licensing first, the system would have moved closer to: use it now, and make the creator stop you later.
After major backlash, the government said it no longer had a preferred option. In plain language, that means the plan hit a wall and the government pulled back.
Why it failed
The proposal did not just upset artists. It ran straight into the way the copyright world already works.
1. It clashed with the basic rule
Copyright is built around permission, licensing, and legal control. The proposal moved in the opposite direction.
2. It shifted the burden
Instead of companies proving they had rights to use the work, creators would have had to monitor and fight back.
3. It ignored the size of the system
Copyright is not just a principle. It is a large business with money, contracts, royalty systems, and legal enforcement behind it.
The copyright industry already has a machine behind it
To understand why this UK plan struggled, you have to understand what it was up against. Copyright is not just a legal idea creators talk about when something goes wrong. It is a working industry.
Who is involved?
- Creators, writers, artists, producers, composers
- Publishers, labels, studios, and media companies
- Collection societies and rights groups
- Law firms and in-house legal teams
- Technology companies that track usage and infringement
- Courts, tribunals, and regulators
What keeps it running?
- Licensing deals
- Royalties and collections
- Catalog ownership
- Monitoring tools
- Takedowns and claims
- Lawsuits when needed
In other words: the fight over AI training data is not happening in an empty field. It is happening inside a system that already makes a lot of money protecting, licensing, and enforcing human-created work.
There is serious money behind copyright
This matters because policy fights usually get harder, not easier, when they touch a market that is already worth billions.
Recorded music
$31.7B
Global recorded music revenue in 2025.
Creator collections
€13.97B
Royalties collected worldwide for creators in 2024.
ASCAP
$1.945B
Revenue collected by ASCAP in calendar 2025.
Once you see the scale, the UK backlash makes more sense. This was never going to be treated like a small technical adjustment.
Before We Go Further
If you want to understand how AI can actually help you, start with the basics.
This issue gets noisy fast. These two links give readers a clean next step: practical AI guidance and a way to stay connected as the conversation keeps changing.
Copyright enforcement was already active before AI entered the picture
One reason the UK plan felt unrealistic to many people is simple: infringement disputes already happen all the time.
United States
The U.S. Copyright Claims Board exists because copyright disputes are common enough to justify a streamlined forum for smaller cases instead of sending everything to federal court.
By March 2025, that board had already received 1,222 claims since launch. In federal court, damages in copyright cases can go much higher, including statutory damages that can reach $150,000 per work for willful infringement.
Canada
In Canada, the Federal Court’s 2024 statistics listed 73 new copyright proceedings and 169 pending copyright matters at year-end.
That is not a picture of a dormant rights system. It is a picture of a rights system that is already used and defended.
The key point is not that every creator sues. The key point is that the legal infrastructure is already there, and it is already used. AI did not invent copyright enforcement. It is running into it.
How the UK compares with the United States and Canada
| Region | What is happening now | Main driver | What that means |
|---|---|---|---|
| UK | The government backed away from its opt-out proposal and says it has no preferred option. | Government policy | Big reset. No final answer yet. |
| United States | Courts continue to hear AI training cases, while the U.S. government has now formally supported OpenAI’s fair-use position in the New York Times litigation. | Litigation + federal legal/policy position | The government is advocating a position; courts still decide the law in the cases before them. |
| Canada | Canada still has no AI-specific copyright framework, but SOCAN has filed a Federal Court action against Suno under existing law. | Existing law + litigation | No AI-specific rule yet, but Canada is no longer only in a study-and-watch phase. |
Why the United States matters so much here
If the UK story was mostly about a policy proposal collapsing, the U.S. story is still about judges deciding real disputes—but the federal government is no longer standing outside the argument.
On September 2, 2026, the U.S. government filed a brief backing OpenAI’s fair-use position in its dispute with The New York Times. Reuters reported that the government argued restrictive interpretations of copyright could harm scientific progress, national security and economic development. That is a significant policy signal, but it does not establish that AI training is fair use as a matter of law in every case.
The same day, Commerce Secretary Howard Lutnick urged G20 countries to develop frameworks that allow AI companies to train on copyrighted works while balancing creator protections. Again, this is advocacy about policy direction—not a judicial decision.
In the U.S., courts have already started drawing lines around AI training. One important theme has been this: using lawfully obtained material for training may be treated differently from using pirated material. At the same time, new lawsuits keep arriving from book publishers, music rights companies, and reference publishers.
On top of that, the U.S. Supreme Court recently declined to take the Thaler case, leaving in place the rule that copyright protection still starts with human authorship, not fully autonomous machine creation.
Canada is now testing AI music under existing copyright law
Canada still has not created a clear AI-specific copyright framework. But the country is no longer accurately described as only studying the issue.
On September 2, 2026, SOCAN filed an action against Suno in the Federal Court of Canada alleging infringement of performing rights in musical works. SOCAN’s claim cites 150 publicly available Suno outputs as examples. The allegations have not been proven in court.
The distinction from the United States matters. Canada uses a fair-dealing framework rather than U.S.-style fair use, so an American fair-use argument cannot simply be transferred north of the border unchanged.
For the detailed music-specific analysis, including the output examples, jurisdiction distinction and what the filing does and does not mean for Suno users, see the JR Suno & Udio Copyright Lawsuits tracker.
Where human contribution still matters
One reason this whole debate feels so one-sided to many creators is that copyright law has long centered human contribution. That does not mean every country uses the exact same wording or gets every AI question answered the same way. But the broad pattern is still there.
In the U.S., that principle is very clear. The Copyright Office and the courts continue to treat human authorship as the baseline requirement for copyright protection. In Canada, authorship questions around AI remain open, but the legal system still treats creators’ economic and moral rights seriously. The UK has its own special rules for some computer-generated works, but the current fight is less about whether software can assist creation and more about whether companies should be able to ingest protected material without permission.
That is a big difference. The public conversation often blurs together two separate issues: who gets copyright in outputs and who had the right to use the inputs. The UK fight was mainly about the inputs.
Why creators should care, even if they are tired of the noise
It is easy to get lost in the online culture war around AI. But beneath all the slogans, the real issue is practical.
Control
Who decides whether your work can be used to train a system?
Compensation
If your work helps build a profitable product, is there a licensing path or not?
Burden
Does the company need to prove it had rights, or does the creator need to chase the company after the fact?
That is why this story matters. Not because every argument online is smart. Not because every critic is fair. Not because every AI defender is honest. It matters because the legal and economic questions underneath it are real.
If you’ve made it this far, you already understand more than most people talking about this topic.
This debate is bigger than “AI slop”
Low-quality AI content is a real problem. So is low-quality human content. But that is not the legal question at the center of this story.
The real question is whether a government can loosen copyright protections for large-scale AI training in a world where copyright is already commercialized, defended, and deeply wired into how creative industries make money.
Once you frame it that way, the UK backlash looks much less surprising.
Bottom line
The UK did not settle AI copyright law. It backed away from one proposal.
In the United States, courts are still deciding the hardest questions, but the federal government has now clearly signalled support for fair-use-based AI training arguments. In Canada, there is still no AI-specific copyright statute, but SOCAN’s Suno action means existing Canadian law is now being tested directly against AI music.
The direction is increasingly jurisdiction-specific: government policy, court decisions, licensing and existing copyright doctrines are moving at different speeds. A headline from one country should not be treated as a universal rule for creators everywhere.
Next Steps
Understand the system. Then use it.
This article breaks down how AI, copyright, and enforcement are evolving. The next step is learning how AI can actually help you and staying connected as the landscape keeps shifting.
No hype. No noise. Just clear systems, tools, and real-world application.
FAQ
Did the UK legalize AI training on copyrighted work?
No. The UK government backed away from a proposal that would have made that easier under an opt-out model. It did not settle the issue.
Is the United States more pro-AI than the UK?
Not in a simple way. The U.S. government has now backed a fair-use position in one major AI copyright case, but courts—not the executive branch—still decide the disputes before them.
Has Canada made a clear AI copyright rule yet?
No. Canada still has no clear AI-specific copyright framework. However, SOCAN’s September 2 action against Suno means existing Canadian copyright law is now being tested directly against an AI music generator.
Is this mainly a debate about low-quality AI content?
No. That is part of the public conversation, but the core legal fight is about copyright inputs, permission, licensing, enforcement, and compensation.
