By Mark Seavy

With the European Parliament’s recent approval of the Artificial Intelligence (AI) Act, guardrails for protecting trademarks and copyrights are being established.

The new law, which is expected to be approved by European Union members by May, requires safeguards that would bar AI systems from generating illegal content. And with the White House having formed an AI Council, similar legislation will likely pick up speed in the U.S. amid ongoing court battles about technology’s use of content and copyrights, according to attorneys at the Intellectual Property Association’s recent The Business of Artificial Intelligence conference in New York.

These regulations will become increasingly important as more companies incorporate AI into their content creation processes.

AI developer Midjourney, for example, has built image and video generators. The company maintains a list of thousands of artists—including illustrators and designers at Hasbro and Nintendo—to help train its AI technology.

However, just because artists are involved in the development of these programs, many governing bodies still require human involvement in the creation of the art itself in order for the content to qualify for copyright protection.

The U.S. Copyright Office has previously barred AI works from being copyrighted if they were created without human intervention or involvement. The Office, for example, rejected artist Ankit Sahni’s request to register his two-dimensional artwork, Suryast, because he identified the Raghav Artificial Intelligence Painting app as the author of the work. The artwork combined an original photo with a copy of artist Vincent van Gogh’s The Starry Night.

And beyond the lack of a human author, another issue at the heart of many of the cases going forward will be what data from the public domain AI can use to create new works and when a developer must be paid a licensing fee, said Jed Ferdinand, a partner at Meister Seelig & Fein.

“The response from many companies when they are sued is ‘we trained [the AI-powered program] on publicly available data,’ but publicly available does not mean public domain and they are not synonymous,” said Josh Weigensberg, a partner at the law firm Pryor Cashman. “There are some data sets that are publicly available to license and some that are public domain and some that are a mix of both.”

While many law firms and regulatory bodies are seeking to head off the use of AI to create content that may infringe trademarks or copyrights infringe, they are also adopting the technology to perform some tasks, said Dan Englander, a partner at the law firm Kilpatrick Townsend & Stockton. In fact, many organizations are using AI to discover the bad actors that are using artificial intelligence improperly.

The U.S. Patent & Trademark Office (USPTO), World Intellectual Property Organization (WIPO), and the European Union Intellectual Property Organization list around 80 AI-enabled functions covering 20 international jurisdictions, including identifying similar trademark images and words using advanced phonetics and tools for conducting searches for similarity.

“A great use is taking a ton of data and having AI help you find and distill things within it that, otherwise it would take a team of associates or contractors a long time to complete,” Englander said. “But in order for that tool to work effectively, you need to have attorneys trained in it to determine what is good and useful for a case. It doesn’t displace anyone, but it can speed things up and lower costs.”