Australia’s national broadcaster has rejected calls to loosen copyright rules for artificial intelligence companies, warning that an opt-out system would force creators to police the use of their own work after AI developers may already have copied large amounts of online content.
The Australian Broadcasting Corporation told a parliamentary inquiry on October 6 that existing copyright rules remain capable of supporting licensing deals between content owners and AI companies.
Its intervention came as OpenAI, Anthropic and Australia’s creative industries clashed over whether the country should change copyright law to make domestic material easier to use for AI training.
FCRF Launches CP-FRM to Build India’s Next Generation of Fraud Risk Professionals
ABC Rejects Claim That Licensing Is Too Difficult
AI companies have argued that licensing every piece of material used for model training is unrealistic.
OpenAI representative Jeff Bleich told the inquiry that companies could not practically “license the entire internet”, according to ABC reporting.
The national broadcaster pushed back.
Kate Gilchrist, ABC’s head of content and legal operations, argued that companies building commercial AI systems should negotiate for the material they want to use instead of receiving a broad legal right to take it by default.
ABC’s position is that the burden should remain on the AI developer to obtain permission or a licence rather than requiring every journalist, publisher, musician or other rights holder to individually block AI training.
ABC Says Its Material Has Probably Already Been Scraped
The broadcaster also raised a more difficult practical problem.
It believes its publicly available content has probably already been collected, or “scraped”, by AI companies.
Scraping involves automated systems downloading large quantities of material from websites so the information can be analysed or incorporated into datasets.
ABC has previously dealt with requests concerning AI scraping, and its public disclosure log records documents relating specifically to the issue.
The concern is that changing the law now could legitimise behaviour that rights holders believe may already have taken place without their permission.
Australia Is Considering an Opt-Out Model
Leaked Australian government documents revealed in September showed officials considering copyright models designed to make AI training easier.
One proposal would allow AI companies to use online copyrighted material after satisfying certain licensing or payment conditions, unless the copyright owner specifically opted out.
Another approach involved payments through a central organisation or existing industry collecting societies that would distribute compensation to rights holders.
Both approaches would represent an important shift.
Under the current system, a company generally needs a legal basis for copying protected material.
An opt-out approach would instead make some material available by default unless its owner actively objects.
Government Has Ruled Out Broader Copyright Exemption
Australia has already rejected one of the AI industry’s more expansive demands.
The Albanese government has ruled out a broad text-and-data-mining exemption that would automatically protect AI developers when copying copyrighted works for model training.
That means the current debate is narrower.
The question is whether the government should create a structured licensing or opt-out system that gives AI developers greater legal certainty while still providing some rights and compensation to creators.
No final model has been adopted.
Anthropic Has Also Pushed for an Opt-Out Approach
Anthropic has argued for a system providing what it calls greater legal certainty for AI training in Australia.
The Claude developer has accepted that a broad copyright exemption lacks political support but has backed a more conditional arrangement under which rights holders could opt out.
The company argues that AI developers need access to sufficiently large and diverse datasets if Australia wants frontier models to be trained domestically.
It has also linked the issue to investment.
AI companies are planning major data-centre developments in Australia, and access to local training material has become part of the wider policy negotiation over whether the country can become a significant AI-development hub.
Creators Say Opt-Out Reverses Consent
Publishers, musicians, writers and other creative organisations see the same proposal very differently.
Their objection is that an opt-out system changes the default from “ask before using” to “use unless stopped.”
That creates a practical problem for smaller creators.
A large broadcaster may have legal teams capable of monitoring AI companies and enforcing exclusions.
An individual writer, photographer or musician may not know that their work has entered a training dataset until long after the copying occurred.
The ABC told the inquiry that such a system effectively places the compliance burden on the rights holder rather than the company building the model.
Australian Law Currently Gives AI Companies No Broad Scraping Exception
Australia’s existing copyright framework does not contain the kind of broad fair-use doctrine available under US law.
A parliamentary report on AI adoption noted that scraping copyrighted material for model development is unlikely to automatically fall within Australia’s existing fair-dealing exceptions.
Those exceptions apply only to specific purposes such as research, study, criticism, review, parody and satire, and the use must also be considered fair.
Commercial AI training that competes with the original material may therefore face greater legal difficulty.
That legal uncertainty is one reason technology companies are lobbying for reform.
ABC Uses AI but Says Creators’ Rights Must Be Protected
ABC’s position is not anti-AI.
Its own AI principles say the broadcaster intends to use artificial intelligence where it can improve public-media services.
But those principles also require human oversight, transparency and respect for creators’ rights.
Its editorial guidance says generative AI should not be used to create material that replicates or substantially derives from third-party content without permission.
That makes its position in the copyright debate relatively consistent: AI can be used, but copying someone else’s work should not automatically become lawful merely because the copying occurs during model training.
AI Investment Is Increasing Pressure for Legal Certainty
Australia is trying to attract billions of dollars in AI infrastructure investment.
That creates a difficult policy trade-off.
Technology companies argue that legal uncertainty around training data could discourage investment or make it impractical to develop frontier models inside Australia.
Creative industries argue that rewriting copyright rules to attract foreign AI companies would effectively transfer value from Australian creators to technology companies.
The government is therefore trying to balance industrial policy against intellectual-property protection.
This Is Bigger Than One Broadcaster
The dispute could determine how Australian journalism, music, books, photography and other creative works are treated by future AI systems.
If an opt-out framework is adopted, AI developers could gain much broader access to online Australian material.
If existing rules remain largely unchanged, companies may need to negotiate more licensing agreements before using protected content.
Either approach will affect how the economic value created by AI is divided between the companies building models and the people whose work contributes to their training data.
What this means for you
The central question is whether AI companies should need permission before training on copyrighted work, or whether creators should have to opt out after their material is already online. Australia has not decided yet, but the outcome could become an important global test of who bears responsibility for AI training data.
Follow for daily updates on cybercrime, corporate fraud, DFIR, hacking, investigations, and digital forensics
