Can AI And ChatGPT Own Intellectual Property? Legal Considerations For Your Business

AI tools like ChatGPT, image generators and coding assistants are now part of everyday business. They can help your team draft content, ideate new products, and even build features faster.

But there’s a big question many Australian founders and managers are asking: who actually owns the intellectual property (IP) in AI-generated outputs? And can AI itself own anything?

Understanding how Australian IP laws apply to AI will help you protect your brand and innovations, avoid infringement risks, and set clear rules with staff and suppliers. In this guide, we break down what’s currently recognised under Australian law, the risks to watch, and the practical steps you can take to manage IP when using AI in your business.

What Does “Ownership” Mean For AI-Created Works In Australia?

When we talk about IP ownership, we’re usually referring to a few familiar rights:

• Copyright in text, code, images, videos and other original works.
• Patents for inventions (devices, methods, processes) that are new and inventive.
• Registered designs for the visual appearance of products.
• Trade marks for brand identifiers like names and logos.
• Confidential information and trade secrets.

Ownership determines who can legally use, license, sell, or enforce those rights.

With AI in the mix, the key issue is whether the output has enough human input to attract IP protection, and if so, who that owner is (you, your employee, a contractor or someone else). As you’ll see below, Australian law currently treats AI systems as tools, not legal owners.

Can AI Or ChatGPT Be An Author, Inventor Or Designer?

Short answer: not at the moment in Australia.

Copyright (author of literary, artistic and other works)

Australian copyright law protects original works authored by humans. Current authority suggests that purely computer-generated material without human authorship does not attract copyright protection in Australia. In practice, that means:

• If a human makes a meaningful, creative contribution to an AI-assisted work (for example, detailed prompts, selection, arrangement, editing and refinement), copyright can subsist-and the human author (or their employer) can own it.
• If the output is essentially machine-generated with minimal human creativity, there’s a real risk no copyright exists at all.

Why does this matter? If no copyright exists, you can still use the content-but you may find it hard to stop others copying it, and you can’t rely on copyright to enforce exclusivity.

Patents (inventorship)

In Australia, an inventor must be a natural person. Courts have confirmed that an AI system cannot be named as an inventor on a patent application. However, if a human made the inventive contribution (for example, by identifying the problem, directing the approach and interpreting results), they can be the inventor-even if they used AI as a tool.

Registered designs (designer)

Similar principles apply here: the human designer who makes the creative choices will be the author of the design. AI may assist with options, but the person exercising judgment and making the selection is the designer.

Trade marks (brand ownership)

Trade marks are owned by the legal person who applies for them and uses them to distinguish their goods or services. AI can’t own a trade mark, but you can. Protecting your brand remains crucial-AI doesn’t change the fundamentals of brand protection.

Who Owns IP In AI Outputs Used In Your Business?

Ownership often depends on who created the work, under what arrangements, and how much creative input they had. Start by mapping the main ways AI outputs enter your business.

Employees using AI at work

• Generally, IP created by employees in the course of employment is owned by the employer (subject to contract terms). This can cover AI-assisted works where the employee provides the human authorship.
• Ensure your employment agreements and policies clearly state who owns any IP developed using AI tools during employment. This reduces disputes and supports your commercialisation plans.

Contractors, freelancers and agencies

• By default, contractors usually own the IP they create unless there’s a written assignment. If contractors use AI to produce content, code or designs for you, you should secure ownership via an explicit IP Assignment and carefully drafted service terms.
• Be specific about what “deliverables” include, how AI may be used, and who owns any outputs and underlying prompts, models or fine-tuned datasets.

Outputs generated directly by your team using AI tools

• Even where you consider the business to be the “creator,” you’ll still need a human author for copyright to subsist. Encourage your team to capture their creative contributions (e.g. prompt iterations, edits and selections).
• Keep in mind that some AI tool terms restrict use or claim broad licences to user inputs and outputs. Review these terms and ensure they align with your ownership goals.

Prompts, system instructions and fine-tuned data

• Short prompts may not be protected by copyright; longer, original instructions could be. Treat them as confidential information and restrict sharing.
• If you fine-tune a model on proprietary data, consider the ownership and confidentiality of both the training set and the resulting weights or outputs.

Brand assets and trade marks

Whatever you create with or without AI, your brand is still yours to protect. Consider filing to register your trade mark for the business name, logo or product names you intend to use in Australia.

Training Data, Web Scraping And Privacy: Are You Allowed To Use It?

AI systems are only as good as the data behind them. If you’re collecting, licensing or scraping data for training or fine-tuning, be careful about copyright, contract terms and privacy obligations.

Copyright in training data

• Copyright protects original text, images, videos and code. There is no broad “text and data mining” exception in Australia. If you copy substantial parts of copyright material to build datasets, you risk infringement unless you have permission or a valid exception applies.
• Even non-copyright materials may be subject to database rights or contract restrictions (for example, website terms that prohibit scraping or reuse).

If your data sources involve scraping public websites or platforms, it’s worth revisiting whether web scraping is legal in Australia in the way you intend to do it.

Privacy and personal information

• If your dataset contains personal information, the Privacy Act 1988 (Cth) and Australian Privacy Principles likely apply. You’ll need a clear basis for collection, only use data for permitted purposes, and implement robust safeguards.
• If your business collects personal information-say through your app or website-publish a transparent Privacy Policy and ensure any third-party processors are bound by appropriate data handling terms.

Confidential information and trade secrets

• Training with confidential business data can create leakage risks if the tool reuses or exposes that information. Limit who can access datasets, use enterprise-grade tools with strong confidentiality commitments, and avoid entering sensitive information into public models.
• When sharing datasets with partners or vendors, put a Non-Disclosure Agreement in place before any disclosure, and clearly describe permitted uses.

Third-party content and licences

• Always check the licence for any third-party datasets, images, code or model checkpoints. Open-source and creative commons licences can be powerful-but they come with conditions (like attribution or non-commercial use) that you must follow.
• Commercial licences should address permitted training uses, derivative works, redistribution and any attribution requirements.

Commercialising AI Products: Contracts, Trade Marks And Risk

If you’re building or deploying AI features in your product, you’ll want your contracts, policies and brand protection strategy to catch up. A few key areas to consider:

Customer-facing terms

• Set clear rules about how your AI outputs can be used, your service limitations, and what happens if inputs or outputs infringe third-party rights. Strong Terms of Use and product disclaimers can help manage expectations and risk.
• If you distribute software, an EULA can set licence scope, restrictions on reverse engineering, and allocation of IP ownership.

Supplier and platform contracts

• When you rely on third-party AI APIs or model providers, review IP ownership, licence scope, training rights, data usage, confidentiality, warranties and indemnities. Ensure the supplier isn’t claiming ownership of your prompts, outputs or datasets.
• If you engage developers or data specialists, ensure your contracts include robust IP assignment, confidentiality and open-source compliance clauses.

Brand protection and marketing

• File for trade mark protection early for names, logos and taglines connected to your AI product to secure brand exclusivity in Australia.
• Be careful with marketing claims about accuracy or bias. Under the Australian Consumer Law, claims must not be misleading or deceptive, and you should have a basis for performance representations.

Workplace policies and training

• Give staff clear rules on when AI can be used, what information can be entered, and how to handle sensitive customer data. A tailored Generative AI Use Policy helps reduce privacy, IP and confidentiality risks.
• Train your team on prompt hygiene, citation practices, and keeping an audit trail of human input to support authorship and compliance.

When to get advice

If you’re making strategic decisions about ownership, licensing models, or investor readiness, it’s wise to speak with an intellectual property lawyer. The right structure and documents now can save you costly rework later.

Practical Steps: An IP And Compliance Checklist For Australian Businesses

Use this checklist to set a sensible baseline for your AI use. Tailor it to your specific model, datasets and go-to-market plans.

1) Capture human authorship and creative input

• Ask your team to document meaningful human input (prompt crafting, selection, editing, arrangement) wherever AI is used. This strengthens the case for copyright protection.
• Encourage final human review and curation for anything you’ll publish or commercialise.

2) Lock down ownership in contracts

• Confirm your employment agreements assign IP created in the course of employment to the business.
• For contractors, ensure you have a written IP Assignment covering AI-assisted outputs, deliverables, datasets and any fine-tuned models or weights.
• Include confidentiality clauses and rules for acceptable AI use in all relevant agreements.

3) Manage data lawfully and safely

• Map your data sources and rights. Obtain licences where needed and avoid copying protected material without permission.
• If you collect personal information, maintain a compliant Privacy Policy and appropriate data processing terms with vendors.
• Minimise sensitive inputs to public models. Use enterprise tools with strong confidentiality settings for any non-public data.

4) Protect your brand and assets

• File applications to register your trade mark for names and logos you plan to use at scale.
• Keep prompts, system instructions and “secret sauce” in a secure repository with access controls. Use a Non-Disclosure Agreement when sharing with partners.

5) Set customer and supplier expectations in writing

• Publish clear Terms of Use, acceptable use rules, disclaimers and IP ownership clauses for your product.
• Negotiate supplier terms that respect your ownership of outputs and datasets, limit retraining on your data, and include appropriate warranties and indemnities.

6) Standardise safe AI use across your workforce

• Roll out a practical Generative AI Use Policy covering permitted tools, approved data sources, and dos and don’ts for sharing information.
• Offer short training sessions on confidentiality, copyright and prompt practices.

7) Keep an eye on legal developments

• AI law is evolving. Monitor updates on copyright, text and data mining, and liability rules as regulators consider reform.
• Schedule periodic reviews of your contracts, policies and licences so they keep pace with your product roadmap and the law.

Key Takeaways

• In Australia, AI systems can’t own IP-copyright, patent inventorship and design authorship all hinge on human contribution and legal persons owning rights.
• To claim copyright in AI-assisted outputs, ensure there is meaningful human authorship (prompting, selection, editing and curation) and keep a record of that input.
• For contractors and partners, ownership is not automatic-secure it with written agreements, including robust IP assignment and confidentiality terms.
• There’s no broad text and data mining exception in Australia, so handle training data cautiously and respect copyright, licences and privacy obligations.
• Commercialising AI safely requires clear customer terms, sensible supplier contracts, and early brand protection through trade marks.
• Company-wide policies and training help your team use AI responsibly and protect your confidential information and IP.
• As the law evolves, regular reviews with an IP professional will keep your strategy compliant and future-proof.

If you’d like a consultation on IP ownership, AI compliance and contracts for your AI-driven business, you can reach us at 1800 730 617 or team@sprintlaw.com.au for a free, no-obligations chat.