Australia has announced a world-leading artificial intelligence framework that will require AI companies to negotiate and pay for access to creative works before using them to train models, marking the first legislated national AI governance approach of its kind. On July 15, 2026, Prime Minister Anthony Albanese unveiled the framework, which includes establishing an Office of AI within the Department of Prime Minister and Cabinet and introducing Australian Standards for AI designed to protect local creators while allowing the country to benefit from AI innovation.

What Does Australia's New AI Framework Actually Require?

The framework establishes several key protections and operational requirements for AI companies operating in Australia. The government has explicitly ruled out granting technology companies a "text and data mining exemption" to copyright law, which would have allowed them to access content for free to train their models. Instead, when the legislation is enacted in early 2027, AI companies will need to negotiate and pay for access to Australian content and data before using it for training purposes.

Beyond copyright protections, the Australian Standards for AI will set clear operational rules for large data centers. These requirements include a legal obligation for data centers to underwrite their own new power supply, pay their full share of connection costs so that energy bills are not impacted by their operations, reduce power when needed to strengthen the grid, and be as water efficient as possible. The federal government will also work with state and territorial governments to ensure that large data centers are built in appropriate locations with input from local communities.

The government's approach will be considered by National Cabinet in August, with standards expected to be legislated early next year. This represents a significant shift toward unified national rules rather than differing state-by-state requirements, which will provide businesses with a single, nationally consistent set of regulations.

Why Are Creative Industries So Concerned About AI Training?

The push for these protections came after months of intense lobbying from technology companies and pushback from creative industries. The Productivity Commission faced significant backlash after its interim report suggested that technology companies should be granted an exemption to copyright laws, effectively giving them free access to content for AI training. Prime Minister Albanese made his position clear by stating that Australia "can be much more than a data warehouse for AI products made overseas" and that Australian "writers, musicians, artists and journalists must retain ownership and control of their work".

The concern is not merely theoretical. Reports indicated that an industry proposal was presented to Cabinet in which AI companies would have been granted special exemptions to mine creative content in exchange for establishing an AUD 350 million per year fund for artists and committing over AUD 50 billion in data center investment. Independent Senator David Pocock called this proposal the "ultimate dirty deal," highlighting the controversy surrounding such arrangements.

How to Understand the Practical Challenges of Protecting Existing AI Models

While Australia's framework represents a significant policy shift, a practical challenge remains: the most advanced AI models from companies like OpenAI, Anthropic, and Google were already trained on vast quantities of publicly available content, including Australian creative works, before any legislative framework existed. Once copyrighted material has been ingested into a model, its influence becomes embedded across billions of model parameters and cannot simply be extracted.

• The "Train Now, Pay Later" Model: The recent Anthropic settlement in the United States illustrates this challenge. Anthropic agreed to pay USD 1.5 billion to resolve copyright claims by authors, but was not required to delete or modify its Claude models. This effectively treats the payment as a retroactive licensing fee rather than a remedy that undoes the original use, creating what commentators have described as a "train now, pay later" model for AI companies.
• Limited Retroactive Effect: Australia's proposed framework will likely have real force for future training runs, as models are regularly retrained and updated. However, for the current generation of frontier AI models, the ingestion of Australian content will have largely already occurred, meaning the legislation's value may lie less in reversing past conduct and more in establishing the terms on which future models must operate.
• Competitive Implications: Lawmakers will need to consider whether a legislative change will further entrench the early movers who have already trained without constraint, potentially disadvantaging new AI companies and competition in the market.

Could This Spark Similar Changes in New Zealand and Beyond?

Australia's announcement presents a significant question for neighboring jurisdictions. New Zealand's Cabinet has asked Minister Brewer to report by March 31, 2027, on a possible copyright framework for generative AI in New Zealand. While Australia is legislating with urgency, New Zealand's approach to date has been deliberately light-touch, with no AI-specific legislation and reliance on existing laws.

New Zealand writers, musicians, journalists, and artists face the same risks as their Australian counterparts regarding unauthorized use of their work in AI training. As Australia moves to legislate, it will be interesting to see whether it prompts a re-examination of New Zealand's current legislative settings. However, New Zealand is actively positioning itself as a destination for data center investment with ambitious governmental targets for foreign investment in AI and data center infrastructure. This creates a tension: mirroring Australia's approach could reassure creators but may deter the very investment New Zealand is seeking.

The broader implication is that Australia's framework could set a precedent for the Asia-Pacific region. If the legislation proves effective in protecting creators while still allowing AI innovation, other countries may follow suit. For businesses deploying or investing in AI across the region, Australia's framework will likely shape expectations and regulatory approaches in neighboring countries.