Copyright Infringement + LLM Lawsuits: Idea-Expression Relief

CVS Health and Asembia stand accused of a massive, years-long copyright infringement scheme, illegally integrating the MMAS tools into their platform without authorization. For nearly a decade, they allegedly used a pirated version of these validated tools to provide diagnoses and treatment recommendations to patients—violating both copyright law and patient trust. This wasn’t a one-time mistake; it was a deliberate, daily operation amounting to millions of unauthorized assessments. With statutory damages of up to $30K per violation, or $150k for willful infringement, the financial consequences could be catastrophic. This case highlights the blatant corporate theft and reckless disregard for intellectual property rights in healthcare. Adherence v. CVS Health Corp. United States District Court, District of Nevada 2:24-cv-01590-JCM-NJK

I'm interested in the situation in which the AI generates the initial image, but then the user uses a computer program like Photoshop to tweak the image, clean up AI artifacts, etc. 1. Is the computer-generated/user-tweaked image suitable for copyright protection? I'm thinking of something done to the AI image with Photoshop or a similar user-operated image manipulation program. 2. Is it a black-line law decision or is it an "it depends" question? 3. In my own expirience, it is impossible to use a given AI program to generate an exact copy of an earlier image using the same prompt, so is there is a significant issue of proof regarding the AI origin of any image. I can see a lot of interesting decisions ahead for the AI system indemnifiers and, of course, judges who dictate or hand-write the first drafts of their decisions.

There is more than idea vs expression at play, say the content providers, because the LLMs injest (copy) copyrighted material. For example, if the output produces a word by word rendition of that material, there is infringement (they say). So, here's where fair use and the Google Books case become the framework. https://en.wikipedia.org/wiki/Authors_Guild,_Inc._v._Google,_Inc. In any event, you are exactly right on the reason for the indemnity program: control the defense. It's a race for precedent on the pivotal question of whether training LLMs on copyrighted material is actionable.

So are we left with copyright infringement if the output image is too close a match to the original? Substantial similarity in output, regardless of how the new image was created. Like, if someone created happy trees wallpaper inspired by Bob Ross, would they be okay? Asking for a friend.

Damien Riehl important Summary thanks for posting!