Home / Technology / AI Admits Limited Copying Amidst Copyright Lawsuits
AI Admits Limited Copying Amidst Copyright Lawsuits
4 Sep
Summary
- Microsoft claims AI rarely reproduces substantial text from training data.
- Analysis of 8.2 million chat logs shows minimal overlap with news content.
- Microsoft argues AI training on copyrighted material is fair use.

Microsoft is contesting copyright claims from publishers and authors by arguing that its AI, Copilot, seldom reproduces substantial text from source materials. In legal filings, the company highlighted an analysis of 8.2 million chat logs, asserting that only a small fraction showed significant word overlap with news content used for AI training. Microsoft contends that using copyrighted material for AI training constitutes fair use, as the resulting AI serves a purpose distinct from the original works.
The filings are part of a consolidated legal battle where plaintiffs allege that Microsoft and OpenAI built products using their copyrighted content and now compete by regurgitating it. Microsoft submitted its arguments as it seeks an early dismissal of the case through summary judgment. If the judge rules in favor of the publishers and authors, the litigation will proceed.