Why the UK Rejected an AI Copyright Opt-Out While Australia Is Considering One

Key Points: Britain originally considered an opt-out copyright exception for AI training but abandoned it as its preferred approach after strong opposition and unresolved evidence questions. Australia, meanwhile, is considering opt-out-based proposals as it weighs creator rights against the economic case for attracting AI training, investment, and infrastructure.

The United Kingdom and Australia have arrived at an unusual point in the global debate over artificial intelligence and copyright.

Britain spent much of 2025 considering a system under which AI companies could use copyrighted material for training unless rights holders explicitly reserved their rights. By March 2026, however, the UK government had stepped away from that proposal as its preferred policy after strong criticism from creative industries and uncertainty about whether the system would actually work.

Australia appeared to be moving in the opposite direction in September 2026. Leaked consultation documents showed that the Attorney-General's Department had discussed options under which online works could potentially be available for AI training by default unless creators opted out. Those ideas have not become Australian law, and the government says consultations are continuing.

At first glance, this looks contradictory. In reality, the two countries are confronting the same problem from different starting points: how much legal certainty should governments give AI developers without shifting excessive costs and responsibility onto creators?

1. Britain Actually Started With an Opt-Out Proposal

Key Point: The UK did not simply defend its existing copyright system from the beginning. Its government originally preferred an opt-out exception.

Britain's December 2024 copyright and AI consultation proposed a broad text-and-data-mining framework that would allow AI developers to use copyrighted works when rights holders had not reserved their rights.

The idea was designed to address a genuine problem for AI developers. Training large models can involve enormous numbers of copyrighted works, and negotiating individual licenses for every item can be expensive or practically impossible.

The government hoped an opt-out system could produce a middle ground: AI companies would receive greater legal certainty while authors, publishers, musicians, photographers, and other rights holders would retain the ability to prevent use of their material.

But the consultation produced substantial resistance. The government's March 2026 report said most respondents rejected the originally preferred broad exception with opt-out. Creative-industry respondents argued that it could allow generative AI systems to learn from their works without compensation and then potentially compete with them.

2. The UK Became Skeptical That Opt-Out Would Work in Practice

Key Point: One of Britain's biggest concerns was not merely copyright principle but whether millions of rights holders could realistically exercise an opt-out.

An opt-out rule sounds simple on paper. A creator who does not want an AI company using a work simply says no.

At internet scale, things become considerably messier, because apparently humanity needed copyright law to become a distributed computing problem too.

A workable system would need reliable technical methods for identifying works, communicating reservations, determining whether AI crawlers respected those reservations, providing transparency about training datasets, and enforcing rights when companies did not comply.

The UK government concluded that significant uncertainties remained about whether technological measures could support an effective opt-out system. It also said there were gaps in the evidence about the broader economic effects of copyright reform. Consequently, a broad opt-out exception was no longer its preferred path.

The House of Lords Communications and Digital Committee had also advocated a licensing-first approach, arguing that reforms should preserve incentives for AI developers and rights holders to reach licensing agreements.

3. Australia Is Under Different Economic Pressure

Key Point: Australia's debate includes a strong concern that existing copyright rules could make domestic AI model training less attractive.

Australia had previously taken a relatively firm position on creator rights. In July 2026, Attorney-General Michelle Rowland said the government had ruled out a general text-and-data-mining exception and emphasized that creators should retain control and receive payment for their work.

Australia's Productivity Commission had also cautioned against rushing to rewrite copyright law, noting that existing licensing arrangements already compensate creators and that AI-related copyright exceptions in other countries were still being tested through litigation.

At the same time, Australian policymakers have heard another argument: if local copyright rules make large-scale training legally difficult, companies may simply train models somewhere else.

The Productivity Commission has previously reported concerns that Australia's copyright framework could discourage certain AI training activities inside the country. That matters because governments increasingly view domestic AI infrastructure and computing capacity as economic and strategic assets rather than ordinary technology investments.

Major AI companies have also been discussing Australian investment. The emerging policy question is therefore not just whether creators deserve copyright protection. Australian officials are also considering whether the legal framework can support domestic AI development without sacrificing those protections.

4. Australia's Opt-Out Proposal Is More Complicated Than “Free Training”

Key Point: The leaked Australian proposals combine opt-out provisions with possible collective licensing or payment arrangements.

Documents reported by Australia's ABC in September 2026 described two options being discussed with rights-holder groups.

Under the proposals, creators would have the ability to opt out of having their works included in certain AI training arrangements. But instead of requiring an AI developer to negotiate separately with every copyright owner, the system could use collective licensing or a centralized payment mechanism.

One version discussed payments through a central body that could distribute money to registered rights holders. Another contemplated agreements with organizations representing groups such as writers or musicians.

That would effectively shift part of the system from individual permission toward collective rights management.

Critics argue this reverses the existing burden. Under ordinary copyright principles, the user generally needs permission unless an exception applies. Under an opt-out structure, the creator may instead need to take affirmative steps to prevent use.

Supporters of reform can counter that individually licensing billions of online works may make large-scale AI training economically or administratively unrealistic.

Most importantly, these remain proposals under consideration. Australia had not enacted this opt-out structure as of September 17, 2026.

5. Why the Two Countries Currently Look So Different

Key Point: Britain has already gone through the political and practical backlash to an opt-out proposal, while Australia is now debating versions of that trade-off.

The clearest explanation is timing.

The UK government proposed an opt-out model, consulted widely, received strong opposition, examined the evidence, and then withdrew that model as its preferred option. It has not declared the copyright debate permanently closed, however. The government continues to investigate alternative approaches and says it wants both a competitive domestic AI sector and a system in which creators can control and receive fair payment for their work.

Australia is currently closer to the experimental stage Britain occupied earlier.

Australian officials are weighing the economic benefits of local AI investment and sovereign computing capacity against concerns that weakening permission requirements could reduce creators' bargaining power.

That pressure became especially visible in September 2026 as leaked policy documents emerged while senior AI industry executives were meeting Australian officials. The government has described the discussions as consultations rather than a final policy decision.

The difference is therefore not simply that the UK values creators while Australia values technology companies. Both governments publicly say they want AI innovation and creator compensation.

What differs is where each country currently draws the procedural line between permission, licensing, collective payment, and the burden placed on rights holders.

Key Takeaways at a Glance

1. The UK originally preferred a broad AI-training copyright exception with an opt-out, but abandoned it as its preferred option in March 2026.

2. British creative industries argued that opt-out would shift too much enforcement responsibility onto individual rights holders.

3. Australia previously rejected a conventional text-and-data-mining exception, but leaked September 2026 proposals show officials considering different opt-out and collective-licensing structures.

4. Australia's debate is also influenced by the desire to attract domestic AI development and infrastructure investment.

5. Australia has not adopted the leaked proposals, and Britain's own long-term copyright framework is still under development.

Issue	United Kingdom	Australia
Opt-Out Model	No longer government's preferred broad approach	Being considered in leaked consultation proposals
Current Status	Further evidence and alternatives under review	Consultation, no final adoption
Creator Concern	Opt-out burden and uncompensated training	Control, compensation and enforceability
AI Industry Concern	Legal certainty and access to training data	Domestic training feasibility and investment
Possible Direction	Greater emphasis on licensing and transparency	Possible collective licensing plus opt-out mechanisms

The Bigger Question Is Who Carries the Burden

The UK-Australia comparison reveals the central problem behind AI copyright policy.

Almost everyone involved claims to support both innovation and creators. The difficult part is deciding who must take action first.

Under a licensing-first model, AI developers generally carry the burden of securing permission or finding legally licensed material. Under an opt-out model, much of that burden can shift toward creators, who must identify themselves, register objections, use technical reservations, or monitor whether their material has been used.

Britain's experience made policymakers more cautious about shifting that burden before reliable technical systems, transparency requirements, and evidence of economic benefits were in place.

Australia is now testing whether collective licensing and payment systems might make an opt-out approach more workable while still attracting AI investment.

Neither country has permanently solved the issue. The important distinction as of September 2026 is that Britain has already stepped back from a broad opt-out model after testing it through consultation, while Australia is actively examining versions of an approach that could place more responsibility on copyright owners.

Sources

UK Department for Science, Innovation and Technology, Department for Culture, Media and Sport and Intellectual Property Office, Report on Copyright and Artificial Intelligence, March 18, 2026.

UK Parliament, Written Statement on Copyright and Artificial Intelligence, March 18, 2026.

UK House of Lords Communications and Digital Committee, Government Response to AI and Copyright Report, May 15, 2026.

Australian Attorney-General's Department, Copyright and Artificial Intelligence policy materials and ministerial statements, 2026.

Australian Productivity Commission, AI and Copyright policy commentary, July 2026.

ABC News Australia, reporting on leaked September 2026 copyright consultation proposals.