Who Owns AI-Generated Content? The Legal Question Every Business Using AI Should Be Watching

When Adobe launched Firefly, its AI image generation tool, the company made a point of emphasising that it had been trained ethically on licensed images from Adobe Stock and public domain content. That claim was enough to get many businesses comfortable using Firefly for commercial work. Then came the class-action lawsuit.

Photographers and digital artists filed suit against Adobe alleging that the training data included their work without meaningful consent or fair compensation, and that the licensing terms attached to Adobe Stock images didn't cover the use of those images to train a commercial AI system. The case is still working through the courts, but it has raised questions that every business using AI-generated content needs to be thinking about.

The Copyright Landscape in Australia

In Australia, copyright law currently requires a human author. Under the Copyright Act 1968, copyright protects original works created by people. IP Australia, the government body that administers intellectual property rights, has signalled that purely AI-generated content, without meaningful human creative input, is unlikely to qualify for copyright protection.

What does that mean practically? If you ask an AI tool to generate an image and use it exactly as output, with no human creative modification, you may not own copyright in that image. Neither might anyone else, which puts it in a position similar to public domain content. Whether this creates any practical problem depends on whether a competitor can use the same or similar output, and how courts interpret "meaningful human input" as cases emerge.

This is still developing law. No Australian court has definitively ruled on these questions, and the international cases that are working through US courts will likely inform how Australian law develops. What's clear is that the current uncertainty creates real risks for businesses that treat AI-generated content as equivalent to commissioned creative work.

Three Risks Businesses Should Understand

The first risk is that the AI tool you're using may have been trained on copyrighted work without permission. Even if the tool is provided by a major company with assurances of ethical training data, the lawsuits against Adobe, Stability AI, Midjourney, and others demonstrate that these assurances are being contested. If a court finds that training data included copyrighted work and that the AI's output is substantially derived from it, there is a theoretical risk that using that output infringes the original copyright holder's rights. This risk is higher for image generation than for text, and higher for outputs that closely resemble a specific artist's style or a specific original work.

The second risk is that AI-generated text is not unique. The same prompt given to the same model by different users will produce similar outputs. If your business publishes AI-generated content without significant editing, there is a real possibility that another business has published something that reads almost identically. Two businesses using the same marketing copy is an embarrassment. In contexts where originality matters legally, such as trademark applications, contracts, or anything requiring a creative work to be distinguishable from others, AI-generated first drafts need substantial human development before they can be relied upon.

The third risk is tool ownership of output. Some AI tools' terms of service include clauses about ownership of generated content. While most major commercial tools are clear that users own the output, this is worth reading before you build commercial assets on a platform. Terms of service change, and what is clear today may not be in 12 months.

What the Adobe Case Means for Tools You're Already Using

The Adobe class-action has had one immediately useful consequence: it has pushed every major AI image provider to be more explicit about their indemnification policies.

Adobe Firefly now includes commercial indemnification for paid subscribers, meaning Adobe will defend you against copyright infringement claims arising from Firefly-generated content used within the terms of service. Microsoft Designer, built on Azure OpenAI image technology, includes similar indemnification for Microsoft 365 commercial subscribers. Google's ImageFX offers enterprise-level indemnification for Google Workspace users.

These indemnification promises are significant but not unlimited. They typically cover use within the tool's intended commercial purposes and exclude misuse, modification beyond what the tools support, or use in ways that violate the tool's terms. They also don't cover every conceivable copyright scenario, just the claims that the training data issues might give rise to.

For most small businesses using these tools for marketing images, social media graphics, and website visuals, the commercial indemnification from the major platforms provides a reasonable level of protection while the legal landscape settles. Using Midjourney, Stable Diffusion, or less mainstream tools without clear indemnification policies carries more risk, particularly for content where copyright matters.

Practical Steps to Reduce Your Exposure

Stick to tools with clear commercial indemnification policies for commercial work. Adobe Firefly, Microsoft Designer, and Google ImageFX are the most defensible choices for a business that wants to minimise copyright risk with AI-generated images.

Add human creative input. The more human judgement, editing, selection, and modification goes into a final creative asset, the stronger the case that there is a human author and the stronger your claim to ownership. An AI-generated image that you've significantly edited, composited with original photography, or used as a reference rather than a final output is in a much better legal position than raw AI output.

Don't claim AI-generated content as wholly original human work. Beyond the ethical dimension, misrepresenting AI-generated content as original commissioned work creates its own legal and reputational risks as detection tools become more sophisticated.

Keep an eye on developments. The Adobe case and others like it will produce outcomes over the next 12 to 24 months that clarify the legal position significantly. IP Australia is expected to update its guidance as Australian courts begin to engage with these questions. This isn't a reason to avoid AI tools, but it is a reason to stay informed and avoid making irrevocable commitments based on assumptions that the law hasn't yet confirmed.

The honest picture is that AI-generated content sits in a legally unsettled space, but using reputable commercial tools with indemnification policies and applying genuine human creative input puts your business in a reasonable position while the law catches up to the technology.