Microsoft argues its AI Copilot barely lifts text from New York Times, filing shows

Microsoft told a federal court that its AI‑driven Copilot chatbot seldom reproduces even single sentences from articles published by The New York Times or other news outlets, a point it raised in recent legal filings against copyright claims.

The company’s brief cites internal testing that found the model “rarely generates verbatim excerpts” and, when it does, the passages are short and lack the substantive content needed to replace the original work.

Copilot, which is built on large language‑model technology, creates responses by predicting likely word sequences based on patterns learned from a massive corpus of text. That process, Microsoft says, does not involve copying whole passages, but rather synthesising information in a way that differs from the source material.

The filing comes amid a wave of lawsuits filed by major publishers, including The New York Times, accusing Microsoft and its partner OpenAI of infringing copyright by feeding protected articles into their models without permission. The suits argue that the AI’s outputs effectively serve as substitutes for the original reporting.

In its defense, Microsoft points to technical analyses that show a minuscule overlap between the model’s output and the copyrighted text, and it argues that any incidental similarity falls within fair‑use parameters. The company also highlights that the AI’s responses are typically a blend of many sources, not a direct lift from any single article.

Legal experts say the outcome could shape how AI developers handle copyrighted material going forward, potentially prompting stricter licensing agreements or new industry standards for data use.

The litigation is still in its early stages, and both sides have indicated they expect a protracted court battle. Meanwhile, regulators and lawmakers are watching closely as the case may set precedent for the broader debate over AI‑generated content and intellectual‑property rights.