Anthropic urges Australia to consider "conditional approval" for Big Tech to train models on copyrighted works, giving copyright holders a choice to opt out

Content creators will soon have an opt-out choice for their work being used to train AI models.

AI company Anthropic has proposed a "conditional approval" system to the Australian government, allowing Big Tech to train AI models on copyrighted Australian works using an opt-out mechanism for creators. This proposal comes after the government ruled out a blanket text and data mining exemption. Anthropic argues this approach would provide legal certainty and support AI development while still respecting national cultural interests. However, Australian public broadcasters ABC and SBS have voiced strong opposition, advocating for strict media regulations that would require AI companies to compensate content creators and adhere to existing copyright, defamation, and privacy laws. They warn that unregulated AI training could lead to the "cannibalisation" of the Australian news industry and erode information integrity.

The push for an opt-out system signifies a pragmatic approach by AI companies to navigate copyright concerns, acknowledging that blanket exemptions are politically unfeasible. By suggesting an opt-out, Anthropic aims to balance innovation with creator rights, potentially setting a precedent for how AI training data is sourced globally. This move also highlights the growing tension between AI development and existing intellectual property frameworks, forcing policymakers to find new regulatory solutions.

Australian broadcasters, however, see this as an insufficient measure, demanding more robust protections and compensation akin to media regulations. Their warning about "cannibalisation" underscores the economic threat AI poses to traditional news outlets, especially public broadcasters reliant on public funding and advertising. The debate in Australia reflects a broader global challenge: how to ensure AI advancements do not undermine vital information ecosystems while still fostering technological progress.