Anthropic Wants Australia's News to Opt Out of AI Training. ABC and SBS Say That's Backwards.
Anthropic just told an Australian parliamentary inquiry what it wants: the right to train Claude on Australian copyrighted content unless the rights holder actively says no.
The country's two biggest public broadcasters answered within the same hearing — and rejected it outright.
Here's what's actually on the table:
→ Anthropic calls it "conditional approval": AI firms could legally train on copyrighted Australian content by default, while creators retain the right to block specific crawlers using robots.txt files
→ Anthropic made this pitch after the Albanese government already ruled out a blanket text-and-data-mining exemption — this is Anthropic's fallback position, not its first ask
→ SBS rejected the mechanism directly, telling the committee "robots.txt mechanisms are routinely ignored or bypassed," and argued the compliance burden should sit with AI companies, not with publishers chasing an "ever-changing set of training bots"
→ The ABC went further, calling the current setup anti-competitive and warning it would "accelerate the cannibalisation of the Australian news industry"
→ Both broadcasters want AI companies brought under Australia's news bargaining incentive — the same regime that already forces Facebook, Google, and TikTok to pay for news content they distribute
→ SBS said flatly that models like ChatGPT and Claude use news content "without permission, attribution or payment" today
→ Sydney hearings begin next week, with executives from both Anthropic and OpenAI set to appear in person
The disagreement here isn't really about robots.txt as a technical tool — it's about who has to do the work and who bears the risk if the system fails. An opt-out model puts the burden on under-resourced publishers to constantly monitor and update blocking signals against bots that keep changing. A "you must get permission" model puts the burden on AI companies to prove they have the right to the content before they use it. Anthropic obviously prefers the first. ABC and SBS, understandably, want the second.
Australia is shaping up as an early test case here — and whatever gets decided in Canberra over the next few months could become a template regulators in the UK and EU borrow from directly.
Who do you think should carry the compliance burden by default — publishers protecting their content, or AI companies proving they have the right to use it?
#AI #Copyright #Anthropic #MediaPolicy #Australia #TechNews #AIRegulation
— 𝔖𝔞𝔫𝔡𝔢𝔢𝔭 ℜ𝔞𝔦𝔷𝔞