IP Tech Knowledgy

Background hero atmospheric image for Initial AI Cases Skeptical of Copyright Infringement

Initial AI Cases Skeptical of Copyright Infringement

Initial court rulings on artificial intelligence (AI) programs are not immediately finding that AI infringes on the materials used to train their knowledge base.  The early cases are tending to allow the usage when the final AI output does not recite the same wording or images it ingests as part of its learning process.  Judges have tended to deny injunctions seeking to force the AI to stop using the underlying material, and some have asked plaintiffs to try to better support their positions.  

The core of most of the growing number of cases filed has been based on copyright law, asserting that the AI-produced work infringes the copyright of the original material.  Courts seem wary to find that the resulting work produced by the AI process is substantially similar to the original work.  Defendants are relying on the core copyright analysis of “idea vs. expression” and – successfully at the moment — asserting that the AI process reforms how the information is presented and does not regurgitate the original material.  

Plaintiffs have raised additional theories, such as the AI-produced work is an unauthorized derivative work of the original, the AI process intentionally deletes any copy management system or copyright notices from the original, or the AI-generated product falsely implies endorsement or connection with the original.  None of these alternate theories have gained strong support from the few courts that have addressed the matter, although many of the initial rulings have been on motions to dismiss or summary judgment, meaning that there are more facts needed to address in the situations.  

One theory that seems most plausible is that AI gathering and using the material violates the Terms of Use of a website from which the original was taken.  If those Terms of Use specifically prohibit using the content to train artificial intelligence devices, it is possible that the usage would be a breach of contract.  However, even the case law on this theory is still too early on which to base a certain conclusion.  

Moving forward, based on the current analysis, a business developing AI should be sure that its resulting product does not restate the original, but merely ingests the original and then produces material worded differently or with different artistic features that are not substantially the same expression of any part of the original.  Conversely, a business that provides content that an AI bot might pull from should include in its Terms of Use prohibitions on being used as AI training.  

The law on this issue is in its infancy and could change with industry standards, legislation, or a sound, well-reasoned judicial opinion.

Ned T. Himmelrich
410-576-4171 • nhimmelrich@gfrlaw. com