Intellectual Property Challenges for AI-Generated Content

Artificial intelligence (AI) solves problems and generates content using training models and datasets, which can include images, videos, music, and other creative works. Copyright law gives creators the exclusive right to use and license their original works. But when AI generates content, who owns the intellectual property rights?

“That has been a big question for the past couple of years,” says Michael N. Cohen, an intellectual property lawyer at Cohen IP Law Group in Beverly Hills, California. “And it’s pretty much resolved for the most part, though it’s still evolving. Under U.S. law, IP protection resides with a human being, not AI.”

Tech companies and copyright holders should understand their IP rights and legal obligations. For legal advice about intellectual property and AI content, talk to an attorney with experience in intellectual property and technology issues.

Can You Get IP Protections for AI-Generated Content?

Under U.S. copyright law, only a human creator can own copyright. As the U.S. Copyright Office states in policy guidance, “Copyright can protect only material that is the product of human creativity.” Content created exclusively by AI programs lacks human authorship and cannot have copyright protection.

Works created by humans using AI tools may be eligible for IP protection if the human conceived and executed the elements of authorship. However, content solely created by AI does not qualify for copyright.

The same principle holds in patent law: Human inventorship is required for patent protection. “The United States Patent and Trademark Office (USPTO) recently issued revised guidance regarding AI-assisted inventions. It reiterates that while AI can be used as a tool, it cannot be the sole inventor. The invention must emanate from a human,” explains Cohen.

“If the USPTO suspects the invention is AI-derived, they may issue an office action requesting more information to determine how conception of the invention occurred.”

Get Help With Technology Law Matters

Technology law is changing and catching up. For legal help with tech law and regulation, use the Super Lawyers directory to find a technology lawyer near you.

Find a lawyer today

Examples of Copyright Rejection for AI-Generated Works

The U.S. Copyright Office reviews copyright registration applications on a case-by-case basis, carefully assessing whether the submitted work is an original work of human authorship. Examples of rejected copyright cases involving AI include:

• AI-generated illustrations for a graphic novel, although a human authored the text and selected and arranged the images
• Rejected artwork generated by AI and then modified by a human since the human could not identify and disclaim the AI-generated aspects
• Base image taken by the human and styled by AI because the AI was responsible for determining how to interpret the output

If you want to use AI to help you create copyrighted content, you may need to significantly transform the work to claim copyright protections. You should also be able to identify your contributions and how you changed the work from the AI-generated material and disclaim the purely machine-driven content.

Using Protected Datasets for AI Training Models

Before getting to AI-generated content, algorithms “learn” based on input from datasets. Machine learning uses training data to identify patterns and make predictive outputs. Much of the data used to train AI models comes from open source information such as books, articles, and Wikipedia.

However, some datasets include copyrighted material that the rights holders do not authorize for use in AI development. Copyright owners can file a copyright infringement claim against AI developers for the unauthorized use of their intellectual property.

Is AI Modeling With Datasets a Fair Use?

Fair use refers situations when others can use copyrighted material without needing a license from the rights holder. Under IP law, courts use a four-factor test for reviewing fair use:

1. Purpose and character of the use (transformative and educational or commercial)
2. Nature of the original work
3. Amount and substantiality of the work used
4. Effect on the market

Using copyrighted material for AI training is not automatically considered fair use. Claims of fair use are individually assessed based on the specific circumstances. For example, if the AI-driven use is primarily for commercial purposes, it is less likely to qualify as fair use.

Unintentional Copyright Infringement With AI

Unintentional copyright infringement is another area of concern when using artificial intelligence. AI algorithms could generate content that directly copies a copyrighted work without the user’s knowledge.

AI-generated inventions use existing material as their training data. If the training data contains copyrighted works, AI-generated queries could produce outputs that infringe the owner’s copyright without attribution to the source.

To recover damages on a copyright infringement claim, the owner must show the AI program had access to their work and created substantially similar content. “Substantially similar” is a fact-specific determination, based on comparing specific elements and assessing the total feel of the work.

With unintentional copyright infringement, ignorance of the law is no defense. Users of AI are responsible for any potential IP violations if they do not check their use of AI-generated content.

Will Federal Trademark and Copyright Laws Change for AI?

The legal frameworks for copyright, patent, and trademark protections are well-establshed in statute and case law. The legal system applies these existing frameworks to new technologies, including AI innovations. Human authorship and inventorship remain requirements in IP law regardless of advancements in AI technology.

That being said, policymakers have introduced legislation to bolster IP protections in the age of AI. For example, the AI Accountability and Personal Data Protection Act and No FAKES Act, if passed, would provide protections against the use, collection, and appropriation of individuals’ data for AI training without express prior consent.

Some state laws fill in gaps left by federal legislation. For example, under California law, AI providers must be transparent about the training data used to train their AI tools, in part, to help detect AI-generated content.

Get Help From an Intellectual Property Protection Attorney

Copyright can exclude others from using their original creations, including for AI training. AI users also have a responsibility to check their AI-generated content to make sure it does not infringe on copyright protections.

For legal advice on IP rights and AI outputs, contact an attorney with experience in intellectual property law and technology issues in your jurisdiction.