Nicola Lucchi (Universitat Pompeu Fabra – Department of Law) has posted The Invisible Author: Generative AI and the Auditability Gap on SSRN. Here is the abstract:
Without verifiable attribution, copyright operates on a presumption it cannot test. The evidence needed to trace protected expression back to human creative choices is held inside developer infrastructure and systematically withheld from courts, editors, and rightsholders, creating an auditability gap that converts current access arrangements into structural legal protection for AI developers. This comment reframes AI-assisted authorship as an auditability problem in human-AI workflows, distinguishing three evidentiary levels, output, interaction and model-side, and showing that existing instruments including detection tools, disclosure requirements, content provenance standards and AI transparency rules all intervene at the evidential level that the gap forecloses. It then examines how frictionless interaction design produces a cognitive delegation loop that erodes authors’ own capacity to reconstruct their contribution. The paper identifies four requirements for judgment-preserving interaction logs, tamper-evidence, privacy-bounding, signal-to-noise handling and semantic grounding, as open research problems, and asks whether auditability and performance can be treated as compatible design objectives in future generative systems.
Recommended!
To receive new posts from Legal Theory Blog by email, get a free subscription to Legal Theory Stack.
Lawrence Solum