The Pulse
Anthropic Urges Australia to Allow AI Training With Opt-Outs
Anthropic has asked Australia to consider conditional approval for AI companies to train on copyrighted works if creators can opt out. The ABC and SBS oppose the approach, warning that it could weaken protections for journalism and shift mo
AI.info Team · Written with AI ·
Anthropic’s compromise meets broadcaster resistance
Anthropic wants Australia to permit AI training on copyrighted works under conditions that let rights holders opt out. The ABC and SBS are pressing the government in the opposite direction: they want AI companies bound by media rules and required to help fund journalism. Their submissions to a parliamentary inquiry put the dispute over copyrighted Australian material—and who should bear the burden of protecting it—at the centre of the policy debate.
In a submission published ahead of the inquiry’s hearings, Anthropic accepted that the government had ruled out a broad text-and-data-mining exception. Instead, the company proposed “a narrow form of conditional approval” for AI training. Anthropic’s submission, quoted by The Guardian
The proposal would allow AI developers to train on copyrighted material while using “do not crawl” instructions in robots.txt files to signal that a website’s content should be excluded. Anthropic said it was open to supporting Australian creators, but the submission’s reported conditions do not amount to a settled licensing or payment scheme. Australia’s government has ruled out the blanket exemption Anthropic says lacks public support; it has not, in this account, endorsed the company’s alternative.
Why the ABC and SBS reject robots.txt
The public broadcasters argue that an opt-out system makes creators responsible for keeping their work away from training bots. SBS warned that robots.txt mechanisms “are routinely ignored or bypassed” and that rights holders would have to keep updating protections as training bots change. SBS’s parliamentary submission, quoted by The Guardian
The ABC framed the dispute as both a copyright issue and a threat to the economics of news. “This unequal treatment will accelerate the cannibalisation of the Australian news industry,” it said, arguing that AI-based services receive more favourable treatment than established media publishers. SBS also warned that using news content without permission, attribution or compensation could erode public trust and damage the sustainability of the Australian news industry.
Broadcasters want AI included in news bargaining
The ABC and SBS want AI companies included in Australia’s news bargaining incentive, which requires covered digital platforms to make commercial deals with large media organisations to support journalism. The Guardian reports that platforms including Facebook, Instagram, Google, TikTok and Snapchat have been required to comply. Applying that policy to AI companies would extend the argument beyond whether a crawler respects an opt-out signal: broadcasters are also seeking compensation and rules that reflect the effect AI services may have on news publishing.
The ABC called the current arrangement “anti-competitive,” saying AI services do not follow the same rules as media outlets. Its position links copyright protections to the obligations already imposed on parts of the digital platform sector. The broadcasters’ submissions do not establish that the government will extend the bargaining incentive to AI firms.
Copyright remains a condition of Australia’s AI plans
Anthropic also urged the government to support “frontier model training” and pointed to plans for substantial Australian investment, including data centres. The company argued that AI could lift productivity and help address national challenges, while acknowledging potential risks to national security, critical infrastructure and public safety. Access to Australian news, books, music and other copyrighted material is one of the contested issues in those plans.
The parliamentary committee is due to hold hearings in Sydney the following week, with Anthropic and OpenAI executives expected to attend, according to The Guardian. The submissions leave policymakers with competing tests for any compromise: whether opt-out signals can protect creators in practice, and whether AI companies should carry obligations similar to those placed on media businesses. For now, the government has rejected a broad copyright exemption, while Anthropic’s narrower conditional-approval proposal remains a request—not a policy decision.