API
< Back to situations

Monitor this situation.

[SITUATION] · [QUIET] · [TECHNOLOGY]

2 clusters · 4 sources · 4 days · First seen · Last updated

AI training and copyright legal disputes

Overview

Legal challenges regarding the use of protected content for training generative artificial intelligence are emerging in multiple jurisdictions. In the European Union, the European Court of Justice is set to preside over a case involving Like Company and Google Ireland. The publisher accuses Google of using news content to train its Gemini AI services. The ruling aims to clarify when rights holders must grant permission for training data and the liability of AI services when they reproduce protected material.

In the United States, the Wikimedia Foundation has intervened in litigation involving OpenAI and Microsoft by filing an amicus curiae brief. The foundation proposes a distinction between model training and output generation. It suggests that pre-training on copyrighted material should qualify as fair use if the output is transformative, but argues that licenses should be required when systems use techniques like Retrieval-Augmented Generation to provide direct copies that substitute for the original work.

Entities

Google · Like Company · OpenAI · Microsoft · Gemini

Timeline

  1. [TECHNOLOGY] 2 sources
    Wikimedia Foundation proposes copyright compromise in AI litigation

    The Wikimedia Foundation has filed a legal brief urging courts to distinguish between AI model training and output generation in copyright disputes.

  2. [TECHNOLOGY] 2 sources
    EU Court to Decide Copyright Use in AI Training

    The EU Court of Justice will rule on the Like Company v. Google Ireland case, a pivotal decision regarding whether copyright-protected material can be used to train generative AI models.

Sources

ad-hoc-news.de · di.se · spider-mac.com · thepublic.org