Artificial intelligence is creating new legal risks for businesses as new technologies pave the way for inadvertent copyright infringement.
Belinda Breakspear, a partner in the digital and intellectual property team at McCullough Robertson (McR), said agent AI could generate infringing works without prompting from users, creating further risks.
Agent AI refers to advanced systems that can autonomously plan, reason, and execute complex multi-step tasks with minimal human intervention.
“Companies that develop, deploy or rely on AI and agent AI systems could be exposed to further risks, contrary to Australia’s current copyright landscape, which provides accessible enforcement channels for rights holders,” Belinda said in an article published by McR.
under Copyright Act 1968 (Cth) (Copyright Act), copyright infringement occurs when a person or company uses all or a substantial part of a copyright work, without permission or a relevant defense (such as a fair dealing exception), in a way that infringes the exclusive rights of the copyright owner.
To claim copyright infringement, the complainant must be the author or owner of the work in question, and the work must also be copyrighted. A breach must also meet the following requirements:
- An act of infringement has been committed regarding a substantial portion of the copyrighted work.
- When comparing works, “objective similarities” exist.
- There is a casual nexus that indicates that the infringement occurred as a result of copying, whether done intentionally or unintentionally.
There are three main types of copyright infringement.
- Direct infringement – use all or a substantial part of the work in a manner that violates the exclusive rights of the copyright holder;
- Indirect infringement – Dealing with infringing copies without authorization (e.g. importing infringing material into Australia)
- Privilege violation – permit, encourage, or direct others to infringe; The responsibility lies with the person who facilitated the infringement.
Ms Breakspear said there were “significantly fewer” AI-related copyright infringement cases in Australia.
“Nonetheless, both sides of the AI ecosystem, from providers to adopters, could be exposed,” she said.
“For example, providers (who build or provide AI systems) may be held liable if they fail to take reasonable steps to prevent compromised output.
“Deployers (who use or integrate AI systems) may be held liable if they prompt, generate, store, or rely on infringing output, and both may be held liable if copyrighted material is used to train the AI system.”
Ms Breakspear said liability ultimately depends on the actions of each party and how the AI system stores, processes and reproduces copyrighted material.
“Historically, however, litigation has been kept to a minimum due to cost and complexity. (A) The proposed small claims pathway is intended to make it easier for rights holders to enforce their rights and address low-value infringement issues, meaning that companies that use or provide AI are more likely to be sued in the future,” she said.
Serious infringements or infringements on a commercial scale may constitute a criminal offence.
This includes creating, importing, distributing, or possessing infringing copies generated by AI for commercial gain, or using AI systems to create or distribute those copies.
Below are McR’s tips for businesses to manage AI risk.
- Audit your AI tools. See what AI systems actually do. Can I store training data, reproduce copyrighted content, or produce material similar to someone else’s work? Identify where infringement could realistically occur.
- See who is at risk. Review your contract with your AI provider. Find out who is responsible if your system outputs infringing material and whether there is any indemnity or warranty that covers your training data, output, or misuse.
- Use licensed content Where you need it. If your AI system (or agent AI workflow) relies on third-party content, obtain permission or licenses rather than assuming that your use of the AI system is covered.
- Implement company rules How staff should use AI tools. Monitor what training data is entering the system and spot check the output for potential compromise risks.
- train staff. Give your employees simple guidance (don’t ask AI to “copy” existing work, don’t upload copyrighted material without permission, don’t report suspicious output).
- Focus on reform. With text mining and data mining exemptions removed, the Australian Government has indicated that Australia’s copyright rules may be further strengthened. Keep your policies and agreements current by tracking updates.
You can contact Belinda Breakspear here.
Read McR’s full article here.
