AI-generated code ownership disputes are emerging in Australian software contracts, as enterprises and vendors negotiate intellectual property rights for code produced by frontier models during outsourced development projects.
Tech & Ideas Desk is a contributing writer covering tech and public affairs for The Sydney Times.
AI-generated code ownership disputes are emerging in Australian software contracts, as enterprises and vendors negotiate intellectual property rights for code produced by frontier models during outsourced development projects. The disputes reflect the uncertainty about whether code generated by AI models is owned by the entity that prompted the model, the entity that owns the model, or the entity that commissioned the work, and they are arising in contracts where the parties have not explicitly addressed AI-generated output. The Australian Copyright Council has issued guidance stating that AI-generated code does not qualify for copyright protection under current Australian law, which creates a gap in the legal framework that contracts must address through commercial negotiation rather than statutory default.
The guidance from the Australian Copyright Council follows a similar position taken by courts in the United States and United Kingdom, where AI-generated works have been found to lack the human authorship required for copyright protection. The position means that code generated by frontier models such as GPT-4, Claude, or Gemini does not attract copyright, and it can be freely copied and modified by anyone who has access to it. The absence of copyright protection is a concern for enterprises that invest in AI-assisted software development and want to prevent competitors from copying the resulting code, and it is also a concern for software vendors that use AI tools to accelerate development and want to protect their intellectual property.
Software contracts are beginning to include specific provisions that address AI-generated code, with parties negotiating whether AI-generated output is included in the intellectual property assignment that typically transfers ownership of developed code from the vendor to the client. The negotiation is complicated by the fact that AI-generated code is often a mixture of model output and human modification, and it is not always clear which parts of the codebase are protected by copyright and which parts are not. The Australian Software Industry Association has published a model contract clause that addresses AI-generated output, specifying that all code delivered under the contract is owned by the client regardless of whether it was generated by AI or written by human developers.
The model clause includes representations and warranties that the vendor has the right to use any AI models and tools that were used in the development process, and that the vendor has not infringed third-party intellectual property rights in the AI-generated output. The representations are designed to protect the client from the risk that the AI model was trained on copyrighted code that is reproduced in the output, which could expose the client to infringement claims from the original code owners. The risk is real, because frontier models are trained on large datasets that include code from public repositories, and the models can reproduce that code verbatim in certain circumstances.
The risk of copyright infringement in AI-generated code is the subject of active litigation in the United States, where several class action lawsuits have been filed against OpenAI, GitHub, and Microsoft alleging that GitHub Copilot reproduces copyrighted code in its suggestions. The litigation is ongoing, and the outcome is uncertain, but the risk is influencing how Australian enterprises approach AI-assisted development. Some enterprises are requiring vendors to warrant that they have not used AI tools trained on copyrighted code in development projects, while others are accepting the risk and relying on the vendor's general intellectual property indemnification to cover any infringement claims that arise.
The Australian Copyright Council's guidance does not address the liability of enterprises that use AI-generated code that infringes third-party copyrights, but the general principle is that the user of infringing code is liable for the infringement regardless of whether the code was generated by AI or copied from a copyrighted source. The principle creates a risk for enterprises that use AI code generation tools without understanding the provenance of the training data or the likelihood that the model will reproduce copyrighted code. The risk is manageable with appropriate contract provisions and vendor due diligence, but it requires attention to detail that many enterprises have not yet applied to their AI development processes.
The interaction between open source software licences and AI model training is another area of legal uncertainty that is affecting Australian software contracts. Open source licences such as the GPL require that derivative works be released under the same licence, and there is debate about whether AI-generated code that is derived from open source training data is a derivative work that triggers the licence obligations. The debate is unresolved, and it creates uncertainty for enterprises that use open source AI models or that commission development work using open source models.
The Australian Software Industry Association's model contract clause addresses the open source risk by requiring vendors to disclose the AI models and tools that were used in development, and by requiring the vendor to comply with all applicable open source licence obligations. The disclosure requirement allows the client to assess the open source risk before committing to the contract, and the compliance requirement ensures that the vendor is responsible for any licence violations that arise from the use of open source models or training data. The clause is a practical solution to an unresolved legal problem, and it is likely to become standard in Australian software contracts as AI-assisted development becomes more common. Explore more enterprise software legal analysis at the Tech & Ideas hub
For the Australian Copyright Council's AI guidance, see Australian Copyright Council. The Australian Software Industry Association's model contract resources are at ASIA. Open source licence information is published at Open Source Initiative.
Direct inquiries, corrections, or documentation concerning this dispatch to our editorial newsroom desk.
Australian AI startups are facing a funding contraction in 2026 as venture capital firms tighten due diligence requirements, with pre-revenue companies facing the steepest reduction in available capital.
The global AI chip shortage is easing in 2026 as NVIDIA H100 and Blackwell allocations stabilise, but export controls on advanced GPUs to Southeast Asia are redirecting demand toward alternative suppliers and custom silicon.
AI regulation compliance costs are rising as enterprise legal budgets expand to cover impact assessments, conformity documentation, and ongoing monitoring requirements for high-risk AI deployments.
Sydney-based financial services firms are deploying Anthropic Claude 4 for document analysis and compliance workflows, citing lower hallucination rates than earlier frontier models.