OpenAI and Microsoft Move to Dismiss New York Times Copyright Lawsuit re Training and Outputs of ChatGPT

OpenAI and Microsoft both filed responsive motions to the lawsuit brought by the New York Times, with OpenAI filing a motion to dismiss on Feb. 26 and Microsoft filing a motion in support on March 4.  OpenAI and Microsoft seek dismissal of NYT’s claims, urging that New York Times manipulated OpenAI’s coding to produce the result it complains of, asserting that New York Times slept on its rights by filing in 2023 instead of when training began, and arguing that training AI is fair use of copyrighted works.

On December 27, 2023, New York Times brought suit against OpenAI and Microsoft, asserting that ChatGPT tools allow copyright infringement of New York Times articles.  The New York Times presented examples showing near-exact replication of a New York Times article from a ChatGPT output.

NTY Compl. ¶ 99.  New York Times also asserts that Browse with Bing allowed users to prompt for entire paragraphs of the articles, without linking to or crediting the NYT, and that GPT-4 was able to access Wirecutter recommendations for products without giving users the NYT’s affiliate links. Compl. ¶ 132.  NYT asserted several counts of legal wrongdoing: Copyright Infringement against all defendants; vicarious copyright infringement (a type of liability for the actions of those operating within your control) against Microsoft; contributory copyright infringement (contributing to infringement with actual or constructive knowledge of infringement); removal of copyright management information under the DMCA (removing copyright notices); unfair competition; and trademark dilution by tarnishment (available for impairing the distinctiveness of a famous brand by leading consumers to believe its quality is lower or reputation is worse than it actually is). New York Times seeks damages and an injunction (to stop the alleged infringements), but also seeks “destruction . . . of all GPT . . . models and training sets . . . .”

On Feb 26, 2024, OpenAI responded to the NYT lawsuit by filing a partial motion to dismiss.  OpenAI is seeking dismissal of claims of copyright infringement in training occurring more than three years ago (under the three year limitations period for damages) and states that NYT should have known more than three years ago that NYT articles were part of the GPT dataset, so any discovery rule for this limitation does not apply.  OpenAI also argued for dismissal of all claims of contributory infringement on the grounds that OpenAI lacked knowledge of infringement by users.  Generally claims of contributory infringement require actual or constructive knowledge of infringing activity and materially contributing to or supplying the means to enable infringement by users.  New York Times does not allege that it notified OpenAI of particular acts of infringement that OpenAI failed to remove. However, in past technology cases courts have declined to find that the owners of a system had knowledge of infringement by users if the system was widely used for legitimate, unobjectionable purposes” or “merely be capable of substantial noninfringing uses.”  Sony Corporation of America v. Universal City Studios, Inc., 464 U.S. 417 (1984). 

OpenAI also claims that it should not be liable for removal of copyright management information under the DMCA, which requires an act of removal of copyright notices or similar info that is designed to induce or conceal infringement, because OpenAI asserts that it did not engage in any act of intentional removal of copyright data to facilitate infringement.  OpenAI also asserts that if using snippets were a DMCA violation, then every journalist who used another’s material in their work could be liable.   OpenAI also asserts that the unfair competition claims should be dismissed on the grounds that they are preempted by the Copyright Act.

On March 4, 2024, Microsoft likewise filed its motion to dismiss. Microsoft argues that the New York Times’s complaint is not a realistic use case.  It asserts that “The Times crafted unrealistic prompts to try to coax the GPT-based tools to output snippets of text matching The Times’s content.”  Microsoft supports OpenAI’s arguments that there can be no contributory infringement because generative AI tools are capable of lawful use and Microsoft had no actual or constructive knowledge of infringement.  Moreover, contributory infringement cannot be shown without establishing that someone directly infringed NYT’s copyright, but NYT doesn’t allege that any actual user used the generative AI technology to produce outputs substantially similar to NYT’s articles.  Moreover, Microsoft states that the DMCA and unfair competition claims fail for the same reasons as OpenAI argued.  Microsoft also urges that it will show on summary judgment (it cannot do everything here because on a motion to dismiss all facts in the complaint are taken as true) why generative AI using large language models is a fundamentally transformative fair use, akin to technologies found to not constitute copyright infringement in past technology cases like Google Search.

New York Times is expected to file a response to the motions within 21 days, so by mid-March 2024. If these motions to dismiss are granted, NYT will likely be given the opportunity to amend to allege facts such as actual or constructive knowledge of infringement to support a contributory infringement claim. Moreover, direct copyright claims will likely move forward..