A Hacker News discussion examines whether 'no source code was copied' remains a viable copyright defense as LLM-powered app development lowers barriers to creation. The thread references a dispute between Papermark and Corgi Dataroom, highlighting how traditional copyright assumptions are being challenged in the AI era.

Legal experts point out that nonliteral copying can still constitute infringement, with most copyright cases involving 'striking similarity' rather than exact code matches. Proving substantial similarity requires demonstrating both access to the original work and close resemblance. This shifts focus toward discovering evidence of knowledge transfer, such as developer communications and training data exposure.

The analysis reveals a critical complication: LLM-authored code lacks copyright protection in the US, meaning companies using AI-generated code may have weakened legal standing. Conversely, if competitors' LLMs were trained on or accessed proprietary code, this creates clearer infringement pathways. The FOSS community originally raised alarms about training data concerns.

Courts face novel questions about AI's role in copyright disputes, particularly when both parties use LLM-generated code. The practical implication is that developers can no longer rely solely on literal code copying as a defense, as substantial similarity and access-based arguments gain prominence in an AI-saturated development landscape.