Best practices for mitigating copyright risks in AI-generated content

The U.S. Copyright Office in Washington, D.C., recently held a briefing on the latest developments in copyright law and findings from their issued report on the subject, with a significant focus on the impact of AI. The discussion highlighted the complexities surrounding the copyrightability of material created with the assistance of large language model (LLM) AI chatbots like ChatGPT and Gemini. While the general takeaway from the report is that you cannot copyright the results of a chat prompt, it further clarified that edited and personalized content can be copyrightable in appropriate circumstances. This evolving landscape underscores the importance of understanding how AI is reshaping the work of the Copyright Office and the broader implications for content creation.

The reality is that AI will soon be as ubiquitous as Excel in corporate America. Many companies have already signed up for Microsoft 365 Copilot, a virtual assistant powered by AI that integrates with Microsoft products (including Excel) to increase productivity and efficiency. As this becomes more commonplace, here are four best practices for mitigating copyright infringement when using AI to create content for corporate websites.

Consider an intellectual property (IP) clearance review. Many companies have a process in place for clearing IP that appears on their web pages and tracking who owns the rights to that IP. Posting content online is the sort of activity where an ounce of prevention is worth a pound of cure. Expanding an existing IP clearance policy to encompass AI-generated content or establishing one is an excellent first step.

Be very careful with AI-generated images. Images are extremely susceptible to infringement claims. If you have not already received a cease-and-desist letter seemingly out of the blue for some long-forgotten image that appeared on your website years ago, count yourself among the lucky. Trolling the web for infringing images that were not properly paid for and cited has become a cottage industry (as AI has made this easy). The prevalence of these risks carries over into AI-generated content, where companies often include AI-generated images with their text. Before uploading any image from your AI-generated content, make sure you know where it came from and that you are confident it is not substantially similar to someone else’s image.

Monitor third-party vendors. Many companies outsource their web-content creation to third party vendors who specialize in creating content. This can be a great arrangement, but if a company posts infringing material on its website, it is not going to matter who created it. Now would be a good time to check in with your vendor on what they are doing to avoid infringement, and to review whether your vendor agreement provides for indemnification.

Make it your own. Sometimes AI-generated text can be too close to existing copyrighted text. One way to avoid infringement in that scenario is through a best practice that applies to any use of AI: Treat the results of your prompt like a draft prepared for your review by someone else at your company. Would you publish a summer intern’s draft without reading it first? Of course not. The same goes for AI-generated content. It is a starting point but not a final product.

Later this year, the Copyright Office will issue a report on the use of copious copyrighted material to train LLMs is copyright infringement. My prediction is that the Copyright Office will conclude it is not, likely under the fair use doctrine. This prediction is based on that the widespread use of LLM chatbots, making it hard to imagine the Copyright Office will insist on putting the toothpaste back in the tube.

The evolving landscape of copyright law, particularly in relation to AI-generated content, presents both challenges and opportunities. As AI becomes more widely adopted in corporate America, it is essential for companies to adopt best practices to mitigate copyright infringement risks. By implementing IP clearance reviews, being cautious with AI-generated images, monitoring third-party vendors and personalizing AI-generated content, businesses can navigate this complex terrain effectively. The upcoming report from the Copyright Office will undoubtedly provide further insights, but it is clear that understanding and adapting to these changes is crucial for future content creation.

Ned Sackman is a shareholder and co-chair of Bernstein Shur’s Litigation Practice Group. An intellectual property litigation attorney, he currently serves as the vice chair of the Copyright Committee for the American Intellectual Property Law Association. He can be reached at nsackman@bernsteinshur.com.