The New York Times asked a federal court on Thursday for permission to amend its copyright lawsuit against Microsoft and OpenAI, arguing that Microsoft built a custom supercomputer specifically to train AI models on Times articles and other copyrighted works. The amended complaint, filed June 26, 2026, sharpens the paper's contributory infringement claim against Microsoft and alleges the company's Times-trained models helped lift its market capitalization by $1 trillion in the past year alone.
The motion follows a recent Supreme Court ruling siding with Cox Communications against Sony, which had tried to hold Cox liable as an ISP for music piracy on its network. The decision raised the bar for contributory infringement: plaintiffs must now show a defendant intentionally acted to induce illegal conduct. The Times is rewriting its claim against Microsoft to meet that new standard rather than risk dismissal under the old framing.
The Times also voluntarily dropped two claims — a separate count of contributory copyright infringement and a trademark dilution claim — against all defendants, streamlining the case to its strongest arguments. The paper said it is not seeking additional discovery, so the case schedule should not slip.
“Today, we asked the court for permission to file an amended complaint that further strengthens our case, clarifying our claim of contributory infringement against Microsoft based on new law and new evidence uncovered during discovery.”— Graham James, NYT spokesperson
Key facts
- 01The New York Times moved on June 26, 2026 to amend its 2023 copyright suit against OpenAI and Microsoft, sharpening its contributory infringement claim.
- 02The amendment follows a Supreme Court ruling for Cox Communications against Sony that raised the bar for proving contributory infringement.
- 03NYT alleges Microsoft built a bespoke supercomputer specifically to train models on curated Times articles, helping drive a $1 trillion market cap gain in the past year.
- 04The Times voluntarily dropped two claims — contributory copyright infringement and trademark dilution — to streamline the case.
- 05Evidence includes side-by-side outputs from ChatGPT reproducing Times articles, plus fabricated citations to nonexistent Times stories.
Microsoft pushed back through a spokesperson, calling the amended complaint "a last-ditch effort by the plaintiff to save its claim from unfavorable precedent set in other recent rulings." The Times counters that revising arguments when legal standards change is routine and that neither defendant is prejudiced by the amendment.
The substantive shift is the supercomputer allegation. The original 2023 complaint described Microsoft's infrastructure as generic cloud compute. The amended version alleges the system was tailor-made for training on copyrighted material, with Times articles disproportionately weighted in the training mix because both companies wanted models that could mimic high-quality journalism. By selecting the works and providing the machinery, Microsoft did more than rent servers, the Times argues — it actively enabled the infringement.
The damages theory ties the alleged infringement directly to Microsoft's stock performance, framing Times-trained large language models as a load-bearing component of Copilot and the company's broader AI product line. That trillion-dollar figure is doing a lot of work in the complaint: it converts an abstract copyright dispute into a concrete claim about unjust enrichment at one of the largest companies on Earth.
Evidence from discovery includes side-by-side comparisons of ChatGPT outputs and Times articles, plus screenshots of users prompting the model to reproduce paywalled content by repeatedly asking for the "next paragraph." In other instances, the complaint alleges, the model produced multi-paragraph excerpts unprompted. The Times also points to hallucinations in which Bing Chat fabricated quotes attributed to Moira Forbes, daughter of Steve Forbes, and ChatGPT invented a Times article tying orange juice consumption to non-Hodgkin's lymphoma.
“Our models empower innovation, are trained on publicly available data, and are grounded in fair use.”— Drew Pusateri, OpenAI spokesperson
OpenAI maintains its long-running fair-use defense. Both companies are betting the court agrees that training on copyrighted text is transformative use that does not substitute for the original product. OpenAI has argued ChatGPT is not a replacement for a Times subscription because the material is used differently.
That defense has wobbled in recent litigation. One of the earliest fair-use wins for an AI company was granted specifically because plaintiffs failed to prove market harm — the inverse of what the Times is trying to demonstrate with its substitution evidence and Wirecutter affiliate-revenue claims. Last June, a federal judge sketched out what he believed could be a winning argument against AI training on copyrighted works, suggesting the doctrine remains unsettled at the trial level.
If the Times convinces the court that ChatGPT functions as a substitute rather than a transformation, the remedies on the table are severe. The paper is seeking permanent injunctive relief and damages, and the most aggressive outcome would require Microsoft and OpenAI to wipe affected models and retrain from scratch — a multibillion-dollar setback for both companies' product roadmaps.
The supercomputer framing is the most consequential move here. Until now, the AI copyright fights have largely turned on the legality of scraping and training as abstract acts. Pointing at a specific piece of hardware built for a specific purpose moves the dispute from policy debate to industrial design, and it gives the Times a concrete defendant act to hang its inducement theory on. Whether or not the court allows the amendment, expect every major publisher with a pending claim to study this filing closely — the bespoke-infrastructure argument is portable, and it gets sharper every time a hyperscaler talks publicly about how purpose-built its AI clusters have become.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
