The legal landscape governing AI and copyright has shifted substantially in the first quarter of 2026, and a comprehensive update published this week by Norton Rose Fulbright provides the clearest map yet of where the law now stands — and where it remains dangerously uncertain.

The most definitive ruling came on March 2, when the US Supreme Court denied certiorari in Thaler v. Perlmutter, effectively closing the door on AI authorship claims under current copyright law. Dr. Stephen Thaler's long-running effort to register copyright in a work generated entirely by his 'Creativity Machine' has now been rejected at every level of the federal judiciary. The principle that human authorship is a 'bedrock requirement of copyright,' as the District Court for the District of Columbia put it, has been left standing without Supreme Court modification. For publishers and authors, this is unambiguously good news: AI-generated content remains, for now, in the public domain by default.

The training data question is considerably more unsettled. The Bartz v. Anthropic case, in which a class of authors sued Anthropic for training its Claude model on their books, produced a nuanced ruling: the court found that AI training on copyrighted books constitutes fair use, but that storing pirated copies of those books does not. The case subsequently settled for $1.5 billion — an estimated $3,000 per work — a figure that will recalibrate every publisher's and author's assessment of what their backlist is worth in the AI training market. The parallel Kadrey v. Meta case reached a similar conclusion on the training question, with claims related to pirated 'seeding' copies remaining active.

The OpenAI litigation, meanwhile, is generating important procedural law about the discovery of AI inputs and outputs. A series of rulings from Judge Ona T. Wang in the Southern District of New York have established that ChatGPT conversation logs are discoverable documents subject to litigation preservation obligations — a finding with broad implications for any company that uses generative AI in its operations.

For the publishing industry, the practical takeaway from Norton Rose Fulbright's analysis is that the training data question is not resolved by the Bartz settlement; it has merely been priced. The Disney v. Midjourney case, which concerns AI image generation trained on copyrighted characters, and the ongoing OpenAI MDL will continue to push the boundaries of what 'transformative use' means in the generative AI context. Publishers who have not yet audited their licensing agreements for AI training provisions are operating without a map in terrain that is being redrawn in real time.