OpenAI and Anthropic have asked an Australian parliamentary inquiry to reconsider how copyright law applies to training AI models.
The Commonwealth Government has already ruled out introducing a broad text-and-data-mining exception. In September submissions, however, Anthropic proposed a narrower system of “conditional approval” that could require investment or other support for Australian creators and cultural activity. OpenAI advocated a framework allowing models to learn from publicly available information while providing opportunities for collaboration with rights holders.
No legislative change has occurred. The Joint Select Committee on Artificial Intelligence is expected to report in November, and creators, publishers and collecting societies continue to oppose uncompensated use of protected works. Reuters reported the proposals on 22 September.
The development matters because it reframes the debate. The choice may no longer be presented simply as a blanket exception or the existing law. A conditional regime could instead turn on consent, remuneration, transparency, provenance and enforceable investment commitments.
For Australian lawyers, those details would be decisive. Any new pathway would need to define what “publicly available” means, whether rights holders may opt out, what records model developers must retain and how creators could challenge non-compliant training. Until then, infrastructure investment does not itself confer permission to reproduce protected material.
Source: Australian Government response on copyright and AI.
Leave a Reply