In the Twilight of Copyright

Ed Ryan, a patent attorney with about sixteen years in practice, opens by noting that generative AI has forced questions about ownership into the foreground much sooner than lawyers expected. He reviews the basics: copyright protects creative expression rather than ideas, and for software it covers both the literal code and the non-literal structure of a program, which courts assess through an abstraction, filtration and comparison test.

The core point is that the U.S. Copyright Office treats work generated by AI without human authorship as not copyrightable. During filtration, those machine-generated parts drop out of what an owner can protect, and registrations must disclose AI involvement. Ryan explains why this matters for open-source licensing: the GPL depends on copyright ownership to attach its conditions, so AI-produced pieces fall outside its reach. He also describes how proving which lines a person actually wrote could become a serious problem in litigation.

In the audience questions he covered public-domain treatment of AI output, whether outputs might count as derivative works of training data (still unsettled), copyright in prompts that depend on how creative they are, and infringement when AI is used to reproduce protected material. He closed with a practical takeaway: the more code an AI generates, the less a developer can own, so commercial teams should weigh whether human authorship is worth the cost.