NYT Amends Microsoft Copyright Suit, Alleges Supercomputer Was Built Specifically to Enable Infringement

Since the New York Times moved to amend its copyright complaint against Microsoft and OpenAI on June 26, new details from the filing have clarified exactly what the amended complaint adds and what the NYT has agreed to give up.

What Changed in the Filing

The original 2023 complaint treated Microsoft's supercomputing infrastructure as generic cloud services. The amended complaint, according to Ars Technica's reporting on the court filing, now specifically alleges that Microsoft built a bespoke supercomputing system ranked among the most powerful in the world with the deliberate purpose of enabling OpenAI to train on NYT content without authorization.

Generic cloud hosting implies passive enablement. A purpose-built system designed for a specific training pipeline implies active participation.

Why the Supreme Court Ruling Forced NYT's Hand

The amendment is driven by the Supreme Court's recent decision in the Cox Communications case, where Sony failed to hold Cox liable for music piracy by its users as an internet service provider. The Court set a new standard for contributory copyright infringement: plaintiffs must now prove the defendant intentionally acted to induce the illegal conduct, not merely that they knew about it or provided infrastructure that made it possible.

Under that stricter standard, the NYT's original framing—Microsoft provides computing, OpenAI uses it badly—was legally thin. The new framing—Microsoft designed and built the system specifically to enable OpenAI's training at scale—is an attempt to satisfy the "intentional inducement" bar.

NYT spokesperson Graham James said in a statement provided to Ars Technica: "As we have long alleged, Microsoft actively encouraged OpenAI to steal our copyrighted works."

What NYT Is Dropping

This is not a purely offensive move. As part of the amendment, the NYT agreed to voluntarily dismiss two claims: one count of contributory copyright infringement and one count of trademark dilution against all defendants. The NYT is streamlining, not expanding.

James framed it as "streamlining the case to its most potent arguments." Microsoft would frame it differently.

Microsoft's Position

A Microsoft spokesperson told Ars Technica the company views the amended complaint as "a last-ditch effort by the plaintiff to save its claim from unfavorable precedent set in other recent rulings."

The Cox ruling cut both ways. It gave NYT a roadmap for what intentional inducement looks like, but it also raised the evidentiary bar. Microsoft's argument is that the NYT cannot clear that higher bar and is recasting its theory of the case to avoid the verdict unfavorable precedent would produce on the original claims.

The NYT counters that the amendment is procedurally standard when legal standards shift mid-litigation, and that neither Microsoft nor OpenAI is prejudiced because the NYT is not seeking additional discovery to support the new claims. It says the evidence already gathered in discovery is sufficient.

The Core Allegations Haven't Changed

Stripped of the legal maneuvering, the underlying dispute is unchanged since 2023. The NYT alleges ChatGPT was trained on millions of its articles without authorization, that the model reproduced NYT content verbatim in outputs, that ChatGPT positions itself as a substitute for a subscription, and that Wirecutter writers lost affiliate commissions when ChatGPT summarized their reviews directly rather than linking out.

None of those core allegations are resolved by this amendment. The amendment targets only Microsoft's liability theory, specifically whether Microsoft was a passive infrastructure provider or an active architect of the alleged infringement.

The Open Question

The heavily redacted nature of the filing—reported by Ars Technica—means the specific evidence NYT says it uncovered during discovery about the supercomputer's design and purpose has not been made public. The court's decision on whether to allow the amended complaint, and whether discovery materials supporting the supercomputer allegation are eventually unsealed, will determine whether the "intentional inducement" claim survives to trial or gets dismissed at the pleadings stage.

The district court has not yet ruled on the motion to amend as of June 26, 2026.