How big tech tried to sneak through a plan to rip off content creators for AI training
Image: Markus Winkler @markuswinkler via Unsplash
Australia’s powerful business lobby came up with a plan that would have allowed tech companies to bypass the copyright system entirely and train AI on books and articles without permission from the creators, or paying them a cent for their work.
Digital news outlet Crikey obtained a draft submission produced by the Business Council of Australia that proposed introducing a new legal concept to Australia’s Copyright Act that it calls “non-expressive use” – effectively a backdoor for ripping off content creators.
Crikey reports the submission is dated December 2025, even though the government had ruled out any changes to copyright law that exempted text and data mining over a month earlier. The BCA quietly dropped its proposal for non-expressive use from its final submission.
Current copyright law protects the specific way ideas are conveyed or composed – a work’s “expression” – but not the underlying facts or ideas. The BCA proposed changing the legislation to define mining text or data for “computational analysis” as non-expressive use, falling outside the copyright system and making it lawful for training AI without permission or payment.
Experts say the proposal amounts to little more than a clever way of trying to sneak in a copyright exemption for text and data mining that was previously rejected.
Crikey points out non-expressive use of copyrighted material is considered fair use in countries such as the US, but only if doing so does not infringe on the rights of the copyright holder. The BCA’s proposal does not contain any equivalent safeguards by locating non-expressive use entirely outside the jurisdiction of copyright law.
The Australian government says it has no plans to introduce any text and data mining exceptions to copyright law for AI training, including through a “non-expressive use” amendment.
However, the BCA, whose board includes executives from the some of the biggest tech companies in Australia, appears intent on watering down copyright protection for AI use. Its final submission proposes AI developers should not be held liable when users deliberately reproduce copyrighted material. In practice, AI companies would have no obligation to ensure their systems didn’t duplicate copyrighted content – whether directly prompted by users or not.
The BCA’s final submission also opposes a compulsory licensing scheme for training AI systems on expressive use of data – heightening fears that tech companies remain determined to undermine copyright protections and fair rewards for content creators.
https://www.crikey.com.au/2026/03/12/ai-artificial-intelligence-copyright-business-council-australia/
