Anthropic, OpenAI urge Australia to weaken copyright for training

This dispute reflects a recurring pattern in international AI governance where technology firms seek jurisdiction-specific copyright exceptions to secure operational footholds. Historically, similar dynamics have occurred in Japan (2018), the United Kingdom (2023-2024), and the European Union during AI Act negotiations, where training data exemptions were debated alongside industrial policy goals. In each instance, the central tension involved balancing domestic AI competitiveness against rightsholder protections, often resulting in hybrid frameworks involving opt-out mechanisms or transparency requirements rather than blanket fair use. The Australian case follows this established trajectory but introduces a novel variable: the explicit linkage of copyright reform to physical infrastructure investment as a bargaining chip. Previous disputes typically framed training access as a free speech or innovation imperative; the current approach ties legal permissions directly to capital expenditure commitments. This pattern suggests a maturation of AI lobbying strategies, moving from abstract principles to transactional policy exchanges. It also mirrors earlier debates over data localization laws, where multinational firms negotiated regulatory concessions in exchange for local presence, raising persistent questions about the portability of legal standards across borders.

In late September 2026, a significant policy dispute emerged regarding Australian copyright law and artificial intelligence training, centering on submissions made by Anthropic and OpenAI to the Australian government. According to reporting by Reuters on September 22, 2026, both AI laboratories formally urged Australian officials to relax existing prohibitions that prevent the use of copyrighted creative content for model training without permission [3]. This lobbying effort coincides with the Australian government’s ongoing review of copyright reforms intended to address the technological realities of the AI era, a process currently accepting public and industry submissions as of Q3 2026.

The core of the controversy lies in the specific mechanism proposed by the AI companies and the potential extraterritorial implications identified by critics. According to The Next Web and Superpower Daily, Anthropic and OpenAI have linked their request for legal reform directly to infrastructure commitments, suggesting that favorable copyright rules are necessary to justify domestic data center investments and local model training capabilities [4][5]. The Daily Star further reports that this pressure comes after the Australian government previously rejected broader proposals for no-consent training exemptions, leading the companies to advocate for narrower legal pathways rather than blanket fair use provisions [4][6].

However, this framing has been sharply contested by critics who argue the proposed reforms contain significant loopholes. On September 29, 2026, Inside Story published an investigation detailing these concerns, which was subsequently amplified on Bluesky by user Huia Sue [1]. According to Sue’s analysis of the Inside Story report, the specific opt-in/opt-out system advocated by the AI firms would effectively create a jurisdictional arbitrage opportunity. The criticism alleges that while the reform is presented as enabling domestic training on Australian content, the legal mechanism would actually permit the copying of global material without compensation, provided the technical act of copying occurs within Australian borders [1]. This interpretation suggests that the proposed exemption could function as a conduit for uncompensated use of international intellectual property, leveraging Australian sovereignty to bypass stricter copyright regimes elsewhere.

Anthropic and OpenAI have not publicly rebutted the specific allegation regarding global content laundering in the available sources, though their documented position emphasizes the necessity of legal clarity to support local AI development. According to Reuters, the companies’ submissions frame the issue as a competitive necessity, arguing that maintaining strict bans places Australia at a disadvantage in attracting AI infrastructure capital [3]. The tension, therefore, is between a national interest in securing physical AI infrastructure and the protection of creator rights, complicated by fears that the legal vehicle for achieving the former may inadvertently undermine the latter on a global scale.

The Australian government remains in a deliberative phase, reviewing these conflicting inputs as part of its broader copyright modernization agenda. No final legislative text or regulatory decision has been announced as of the latest available reports. The situation represents a test case for whether infrastructure investment pledges can successfully influence national intellectual property standards, and whether such influence carries unintended consequences for international copyright norms. The discourse highlights a recurring friction in AI policy: the gap between the stated economic benefits of regulatory relaxation and the technical-legal realities of how those regulations interact with borderless digital supply chains.

Australian copyright holders face potential loss of licensing revenue if the alleged loophole enables global content copying through domestic processing. AI laboratories risk losing a strategic Asia-Pacific training hub if Australia maintains strict bans, potentially redirecting data center capital to jurisdictions with more permissive regimes. The Australian government balances competing interests: attracting AI infrastructure investment versus protecting creator rights and avoiding international copyright disputes. Magnitude remains unquantified in available sources—no dollar figures, job counts, or user metrics are cited. The stakes are primarily precedential: outcome could influence how other mid-sized economies negotiate AI training access versus IP protection.