Blog: VLP Speaks

Copyright and AI-Generated Images: What You Need to Know – A Blog Post by Michael Whitener

Posted on Jun 2, 2025 in Artificial Intelligence, Blog, Intellectual Property by Michael L. Whitener

The image-generating capabilities of AI tools such as DALLE-E, Midjourney and Stability AI’s DreamStudio have improved at a remarkable clip. It seems like only yesterday that AI was baffled by the challenge of producing human hands with precisely five fingers. Now, if you want a photo-realistic image of a tiger and a young girl playing chess in Central Park while a giraffe looks on, you only have to ask, and it’s delivered in seconds:

If you’d like to recast that image in the whimsical style of Studio Ghibli, a famed Japanese animation studio, just tweak your prompt:

But what are the copyright implications of creating, using and publishing these AI-generated images? For instance, in the “Ghiblified” image above, might Studio Ghibli founder Hayao Miyazaki have a valid claim that the image infringes Studio Ghibli-owned intellectual property? That issue is now central to a growing number of lawsuits and policy debates.

How AI Tools Are Trained to Generate Images

Let’s start by looking at how today’s AI image generators achieve their striking realism and stylistic fluency. These tools are trained on massive datasets containing billions of images scraped from the internet, tapping into a wide range of sources: public domain works, user-generated content platforms, news sites and digital libraries. Much of this training material consists of copyrighted works created by human artists, photographers and designers – scraped without the creators’ consent.

Developers use this data to teach models how to associate words with visual features and styles. The result: a model that can produce images based on textual prompts by predicting what an image associated with a particular phrase might look like. When a user prompts an AI to create an image “in the style of Studio Ghibli,” it doesn’t pull from a predefined Ghibli database, but rather mimics the style based on patterns it inferred during training.

Can You Copyright an AI Image?

U.S. copyright law requires human authorship. The Copyright Office has consistently denied registration to works that lack “human creative input,” including images generated entirely by AI in response to a user prompt. Consequently, copyright ownership can’t be asserted over such outputs.

That leaves users of AI-generated images in legal limbo: the images typically can’t be copyrighted, and the users could be subject to claims of IP infringement.

What Are the Legal Risks?

Using AI-generated images can expose you to legal claims on several fronts:

• Infringing Third-Party Copyrights. The most significant risk is that your AI-generated image may inadvertently reproduce elements of existing copyrighted works. AI models are trained on billions of images, many protected by copyright. When these systems generate new images, they sometimes produce content that bears striking similarities to images in their training data. If an AI-generated image is found to replicate all or a substantial portion of a copyrighted work, it may be deemed infringing under current copyright law.
• Trademark Violations. Beyond copyright concerns, AI-generated images might inadvertently include logos, characters or distinctive visual elements that are trademarked, exposing a publisher of these images to trademark infringement claims.
• Right of Publicity Issues. If an AI-generated image resembles a real person (especially a celebrity), publication of this image might violate their right of publicity – i.e., their right to control commercial uses of their likeness.
• AI Platform Limitations. Many AI image generation platforms have terms of service that place significant restrictions on how their outputs can be used. Some prohibit certain commercial applications or require attribution. Violating these terms could lead to account termination or even legal action.
• Client Expectations and Contractual Obligations. If you’re creating work for clients, using AI-generated images without proper disclosure could violate contractual obligations, especially if your client agreement includes representations about originality or ownership of the delivered work.

One might hope that the AI tool providers themselves would offer some protection against IP infringement claims, but that’s rarely the case. The terms of use of OpenAI (developer of ChatGPT as well as DALL-E) are typical: they assign to the customer any rights of OpenAI in “output” (including AI-generated images), but pointedly disclaim any warranty of noninfringement regarding that output. OpenAI’s more robust Business Terms for enterprise customers offer some indemnification protection (for example, if the AI services themselves infringe third-party IP rights), but OpenAI stops short of indemnifying for output-related infringement claims.

But Is It “Fair Use”?

One important defense available when AI-generated content is alleged to infringe copyright is the “fair use” doctrine – originally developed by courts and now codified in the U.S. Copyright Act. Under the Act, four factors are weighed to determine whether a use qualifies as fair and therefore non-infringing: (1) the purpose and character of the use (with emphasis on whether it is transformative and/or commercial), (2) the nature of the copyrighted work (creative vs. factual), (3) the amount and substantiality of the portion used, and (4) the effect of the use on the potential market for or value of the original work.

The U.S. Copyright Office recently released a report on AI training that includes a detailed examination of the fair use defense for AI generation. The Office’s conclusion is hardly a ringing endorsement of that defense: “The Office expects that some uses of copyrighted works for generative AI training will qualify as fair use, and some will not.” Ultimately courts will have to decide, absent legislative action on this topic. For now, the question is far from settled.

Mitigation Strategies: Reducing Your Legal Exposure

Despite these risks, there are practical steps you can take to minimize your legal exposure when using AI-generated images:

• Choose Your AI Platform Wisely. Some AI image generators are trained exclusively on licensed content or public domain works. For example, Adobe’s Firefly is trained on Adobe Stock images and public domain content, providing a higher level of assurance against copyright claims. Getty Images and Shutterstock have also launched AI image generators trained only on their licensed content.
• Craft Your Prompts Carefully. Avoid prompts that specifically reference copyrighted characters, artists’ names, or distinctive styles. Instead of “create an image in the style of [Famous Artist],” try describing the aesthetic elements you’re looking for: “create an image with bold colors, strong contrasts, and geometric patterns.”
• Review and Modify. Don’t use AI-generated images “as is” for commercial projects. Instead, use them as a starting point and make substantial modifications to create something truly original. This additional creative input may also strengthen your copyright claim to the final work.
• Maintain Documentation. Keep records of your prompts, the images generated, and any modifications you make. This documentation could prove valuable if questions arise about the origin of your work.
• Consider Obtaining Clearance or Insurance. For high-profile commercial projects, consider seeking legal clearance for AI-generated images or obtaining errors and omissions or other insurance that covers potential IP claims.
• Be Transparent with Clients. Disclose your use of AI tools to clients, and ensure your contracts address the unique copyright situation of AI-generated content. This transparency can prevent misunderstandings and potential disputes later.
• Stay Informed About Legal Developments. Copyright law regarding AI-generated content is evolving rapidly. Current lawsuits involving companies like Stability AI and Midjourney may set important precedents. Regular consultation with an IP attorney can help you navigate this changing landscape.

Finding the Right Balance

AI image generation tools offer tremendous creative possibilities, but they come with legal risks that can’t be ignored. The key is finding the right balance: leveraging these powerful tools while implementing safeguards to protect yourself and respect the rights of others.

The safest approach is to use AI-generated images as inspiration or components within a larger creative process, rather than as finished products. By adding your own creative judgment and substantial modifications, you’re not just reducing legal risks – you’re creating something truly unique that AI alone couldn’t produce.

The world of AI and copyright will continue to evolve, but with thoughtful consideration and appropriate precautions, you can harness the creative potential of AI image generators while keeping legal risks at bay.