Businesses weigh copyright risks from generative AI

A 2025 McKinsey report found that 71% of companies use generative AI for at least one business function, from marketing copy and visuals to coding assistance. That adoption is outpacing legal clarity, especially around whether training AI systems on copyrighted books, articles, images, websites, and other works without permission is lawful.

Courts are applying the Fair Use doctrine to AI training disputes, weighing whether the use is transformative, the nature of the original work, the amount used, and potential market harm. AI developers argue that models learn patterns and create new works, while critics contend that commercial systems depend on large-scale copying and can produce outputs that compete with the originals. No single standard has emerged.

For companies, the more immediate risk often comes from their own AI outputs and inputs. Prompts that ask a tool to imitate a specific artist, author, competitor campaign, or visual style can increase infringement exposure. AI-generated code can also raise open-source licensing issues, while uploading customer information, internal reports, or third-party material may violate copyright licenses or confidentiality obligations.

Businesses can limit risk by adopting written AI policies, approving specific tools, banning prompts involving confidential material or imitation, assigning legal or compliance oversight, reviewing material before publication, and training teams on appropriate use. Strong governance can help companies scale AI use while copyright law remains unsettled.