Protecting IP in the age of AI: a guide for businesses
Whether it is a patented design, formulae, innovation, brand or process that the business uses to manufacture goods and services itself or licenses to others, or creative work from which the artist derives income – intellectual property (IP) is the foundation upon which many businesses are built.
But we have also entered an era of rapidly-advancing artificial intelligence (AI) and the need to train models such as large language models (LLMs or, as they are better known, chatbots). According to OpenAI, its standalone AI chatbot ChatGPT receives 2.5 billion prompts from users on a daily basis – that is a lot of queries and the AI needs lots of data to answer them.
The issue of training AI on copyrighted material is making headlines around the world. It is also an issue that Australian businesses need to address, including how to protect their IP.
IP vs AI
On the one hand, there are the creatives (including designers, artists, writers, photographers, musicians, engineers, academics, scientists, inventors, graphic designers, architects, composers, tech developers) who rely on their IP to make money, to advance their careers, or to simply share knowledge or beauty.
On the other, there is the insatiable appetite of AI companies for content upon which their LLMs can be trained. AI is trained by data being fed into machine learning (ML) algorithms to help them recognise patterns, make predictions, and learn from experience. The AI created is only as good as the content upon which it was trained.
Advanced AI systems can replicate ideas, designs and even code. It is designed to learn from existing data, which can sometimes include copyrighted and other proprietary materials. This means that a business’ IP could be infringed.
Some AI developers argue that all material should be available to them to train their AI models. Many creatives and owners of proprietary data reject this as it infringes their IP and potentially jeopardies their livelihoods.
The World Intellectual Property Organization defines IP as: “creations of the mind, such as inventions; literary and artistic works; designs; and symbols, names and images used in commerce. IP is protected in law by, for example, patents, copyright and trademarks, which enable people to earn recognition or financial benefit from what they invent or create.”
With the rapid advancement of AI, there is growing global scrutiny over how AI companies source their training data and, in particular, the use of IP in such training.
UK allows AI to be trained on copyrighted data
“Betrayed” – that is how music icon Sir Elton John described the UK government’s plan to allow AI to be trained on copyrighted music and creative works without a guarantee of compensation.
In an effort to make the UK a global AI leader, British Prime Minister Keir Starmer proposed allowing AI developers to use whatever content to which they have legal access to train their models. Under the plan, creators would need to actively opt out to prevent AI companies training their models on the creator’s content.
Meta’s former president of global affairs, Sir Nick Clegg, believes asking for permission for every single copyright order would “kill the AI industry” in the UK.
However, former movie director and politician Baroness Beeban Kidron said the government would be “knowingly throwing UK designers, artists, authors, musicians, media and nascent AI companies under the bus” by approving the bill, calling it “state-sanctioned theft” from the UK’s £124 billion creative industry.
Sir Elton was joined by other creatives in condemning the plan to weaken copyright law for AI training. John was just one of 400 creatives including fellow musical artists Dua Lipa, Ed Sheeran, Florence Welch, Kate Bush, Coldplay, Robbie Williams, Annie Lennox, Sir Tom Jones, Eric Clapton, Sting, Mark Ronson and Sir Paul McCartney, to sign a letter calling for the Starmer government to update copyright laws.
The letter urged the PM to back Baroness Beeban Kidron’s amendment to the Data (Use and Access) Bill which would require tech companies to be transparent about the creative content they use to train generative AI models. “This transparency would allow creators and creative businesses to hold AI firms accountable for the mass theft of creative works that continues to take place,” reported In Publishing.
The letter, signed by musicians, writers, producers and actors from across the creative industry in the UK, said that the artists are not against the technology, but stressed that copyright is what allows creatives to own their work and make a living.
The House of Lords rejected the Bill which would make it legal for tech companies to use music, creative works and other copyrighted content they have “legal access” to, to train AI models, four times.
The bill was finally passed by both Houses of Parliament after the government proposed some amendments that improve transparency as to what data is being used. The passing of the bill means AI models can be trained on copyrighted material without the rights holders’ knowledge.
Class action launched for pirating content
Another blow was struck to creatives across the pond.
In June, a US federal judge ruled that Anthropic had not breached copyright laws in training its Claude LLMs, despite not getting consent from the authors of the published books.
Judge William Alsup ruled in favour of Anthropic saying: “The use of the books at issue to train Claude and its precursors was exceedingly transformative and was a fair use under section 107 of the Copyright Act”. (’Fair use’ as outlined in the 1976 legislation is up to interpretation by judges.)
But the creatives have struck back – launching a lawsuit against Anthropic. The Authors Guild said: “the court certified a class comprised of legal and beneficial owners of the rights in copyright-registered books downloaded by Anthropic.” A class action is to follow, representing the copyright owners of the books that were unlawfully downloaded by Anthropic. “If the plaintiffs succeed at trial and show wilful infringement, statutory damages range from $750 per title up to a maximum of $150,000 per title,” stated the Authors Guild.
Other lawsuits have been filed around the globe for copyright infringements by AI companies. One such action was brought against OpenAI, the US company that makes ChatGPT, by news agency Asian News International. The plaintiff alleges that OpenAI used their content without permission to train its chatbots. OpenAI denies the allegations and told the BBC that it uses “publicly available data” that are in line with “widely accepted legal precedents.”
Copyright and AI in Australia
The issue of using IP to train AI models is also being considered in Australia. The complex issues regarding ownership, usage rights, and the legal status of AI-generated content is under scrutiny.
In December 2023, the government established the Copyright and Artificial Intelligence Reference Group (CAIRG) to engage with stakeholders on copyright issues related to AI. CAIRG is focussed on the use of copyright material as inputs for AI systems and aims to develop guidelines for transparency and ethical use of AI in creative processes, notes the Arts Law Centre of Australia.
The Productivity Commission has also recommended that AI companies should have full access to the sum total of Australian artistic output. Using a text and data mining (TDM) exception for the Australian copyright act, it would be legal to train LLMs on copyrighted Australian work without needing permission from the copyright holders. AI training would be added to the list of ‘fair dealing’ exceptions already existing in the copyright act. Musician and former politician Peter Garrett, Arts Minister Tony Burke and others including the Australian Society of Authors, and Writing Australia are against the proposal. Burke stated: “We have copyright laws. We have no plans, no intention, no appetite to be weakening those copyright laws based on this draft report that’s floating around. Use of your work for a commercial purpose, which you have not authorised, is theft.” Following the government’s economic reform roundtable in late August where AI was discussed, it was announced that Australia’s tech sector and unions will work together on a model that pays creatives for content used to train AI.
The Copyright Act 1968 does not explicitly address AI-generated works (of course, AI wasn’t an issue in the late 60s), leading to uncertainties about how existing copyright laws apply to new technologies, the Arts Law Centre said. It also said that, as a result of this gap in legislation, many aspects of AI and copyright remain untested in court.
In addition to issues in relation to the use of copyrighted materials to train AI, there is also the issue of who owns the copyright on materials generated by AI.
Protecting a business’ IP
According to the World Intellectual Property Report 2022, some 61% of businesses using AI reported facing IP-related challenges. A 2025 global study by Arctic Wolf found 45% of Australian and New Zealand businesses identified protection of IP as a top security concern.
Engineering Insights notes IP can be put at risk by AI models due to:
- Data volume and sensitivity: AI models thrive on vast, often complex datasets, and many of these contain sensitive or proprietary information – anything from customer details to unique product designs.
- Open data exchange: training and testing AI models usually involves multiple platforms and systems, sometimes even third-party vendors. Every exchange can open a door to potential data leakage or unauthorised access.
- Model transparency risks: some AI models, especially machine learning models, can inadvertently reveal patterns in proprietary data. For example, when models are reverse-engineered, there is a risk that unique insights or methods derived from proprietary data could be extracted.
For many businesses, IP and AI is an issue that needs to be considered.
One aspect is in protecting the business’ existing IP. Another is in ensuring that the business’ use of AI does not infringe the IP of another.
Tips – Protecting IP in the age of AI
A business’ intellectual property (IP) can be its lifeblood. AI is trained using data that may be proprietary, for example through copyright, patents, or trademarks.
To help protect your IP from infringement, consider:
- Using robust cybersecurity to prevent data breaches and safeguard proprietary information.
- Protecting valuable IP with patents and trademarks.
- Clearly labelling all IP once you’ve registered or been granted an IP right (e.g. with ® registered trademark or © copyright symbols).
- Developing clear agreements and policies regarding ownership rights when collaborating and in contracts.
- Deploying AI-powered IP management systems to detect infringements.
- Keeping up to date on evolving legal standards and regulatory developments.
- Developing a response plan for IP infringements.
When using AI, Engineering Insights suggest you may be able to help protect your IP by:
- Securing your data using data classification and access controls.
- Using data anonymisation (removing personally identifiable information or other proprietary markers from your data before it enters the AI model) and encryption (both in transit and at rest to prevent unauthorised access).
- Conducting regular audits and risk assessments to ensure that your IP protection protocols are keeping up with the evolving AI landscape.
- Leveraging differential privacy techniques which add ‘noise’ to datasets, making it hard for outside parties to reverse-engineer data while still allowing the AI model to generate meaningful insights.
- Defining clear IP and data use policies for AI projects, including IP ownership agreements and data use guidelines.
- Implementing a strong IP protection culture within your business.
There is also the issue of copyright ownership of AI-generated content. This is a complex area and engaging the services of an IP professional is recommended.
Beyond protecting your business’ IP, there is also the risk that the AI tool you have used has copied (or been trained on) someone else’s protected work. “If the output from your AI tool is too similar to another artist’s or company’s content, you could be liable for copyright infringement. This applies even if the copying is unintentional or done by the AI, not you,” notes Sprint Law.
To help safeguard the IP of others, consider:
- Assessing how the business uses AI.
- Identifying areas where AI interacts with IP such as data sourcing and content creation.
- Establishing clear legal frameworks and AI use policies.
- Training employees on AI IP policies.
- Ensuring AI-generated content is carefully checked and vetted for IP infringement.
- Modifying AI-generated content significantly to ensure originality.
- Seeking permissions to use IP where necessary.
- Implementing an AI IP compliance monitoring system.
- Developing a response plan for IP infringements.
IP insurance
IP insurance helps mitigate the financial costs associated with enforcing or defending against IP infringement claims.
This cover can help protect your business against claims of IP infringement. It provides financial protection against risks involving patents, copyrights, trademarks, and trade secrets. It can also help you pursue anyone who infringes on your patent, design right, copyright, trademark or other IP rights.
IP Australia notes a policy typically includes coverage for legal costs to:
- Enforce IP infringement claims– including damages from loss of profits, reputation or settlements.
- Defend IP infringement claims– covers you from claims against you, your customers or licensees. An infringement claim often refers to a violation of a copyright or some other type of IP theft.
Key takeaway
As AI advances, the issue of using IP to train AI models will increasingly be scrutinised and tested in courts around the world.
IP is a valuable asset for a range of creatives and businesses – and protecting those assets is an imperative for many. Businesses are encouraged to look at strategies to ensure they safeguard their IP and also avoid infringing the IP of others when using AI.
Need expert guidance?
Your EBM Account Manager can provide you with information on IP insurance options to help protect your business.
