OpenAI responds to the New York Times

by Stephen M. Walker II, Co-Founder / CEO

OpenAI responds to the NYT

OpenAI has publicly responded to a copyright infringement lawsuit filed by The New York Times (NYT), claiming that the lawsuit is without merit. The lawsuit, which was filed in late December 2023, accuses OpenAI and Microsoft of using the NYT's copyrighted articles without proper permission to train their generative AI models, which the NYT contends constitutes direct copyright infringement.

The team at Klu.ai reviewed the lawsuit, analyzing some of the model's outputs and it appears that the NYTimes used OpenAI's browsing feature to retrieve an article, and then asked the model to repeat the content after retrieval.

Any model with internet access is capable of reading and reproducing any content verbatim that is available on the open web.

Klu Retrieval

In its response, OpenAI has argued that using publicly available internet materials for AI model training is fair use, a principle supported by long-standing and widely accepted precedents.

OpenAI also claims that the NYT "intentionally manipulated" prompts to elicit "plagiaristic" responses from OpenAI's models, suggesting that the examples provided by the NYT were not representative of typical model behavior.

OpenAI maintains that any single data source, including content from the NYT, is not significant for the model's training and that the instances of content regurgitation cited by the NYT were rare bugs.

Furthermore, OpenAI has emphasized its support for journalism and its partnerships with news organizations, stating that it has offered an "opt-out process" to prevent their tools from accessing certain content and that it is continually working to make its systems more resistant to adversarial attacks that could lead to regurgitation of training data.

OpenAI has also highlighted its collaborations with other news outlets, such as the Associated Press and Axel Springer, to provide AI-assisted tools for journalists and explore mutually beneficial opportunities.

The outcome of this lawsuit could have significant implications for the AI industry, digital content creation, the future of journalism, and the boundaries of fair use in the digital age.

What is the legal basis for OpenAI's defense against the New York Times' lawsuit?

The New York Times has not been reported to have faked the results in their lawsuit against OpenAI. The lawsuit alleges that OpenAI and Microsoft used millions of the newspaper's articles without permission to train their AI models, which could potentially infringe on copyright laws.

OpenAI has publicly responded, claiming that the lawsuit is without merit and that training AI models using publicly available data from the web is fair use. The New York Times has provided evidence of instances where the AI models generated outputs that closely mimicked or even verbatim recited content from the Times.

OpenAI has argued that the instances cited by The New York Times appear to be from years-old articles and suggest that the prompts may have been intentionally manipulated to elicit such responses.

The case is complex and involves discussions around fair use, the transformative nature of AI-generated content, and the potential impact on the journalism industry. There have been no credible reports or evidence to suggest that The New York Times faked results to support their lawsuit.

More terms

What is selection in a genetic algorithm?

Selection is the process of choosing individuals from a population to be used as parents for producing offspring in a genetic algorithm. The goal of selection is to increase the fitness of the population by favoring individuals with higher fitness values. There are several methods for performing selection, including tournament selection, roulette wheel selection, and rank-based selection. In tournament selection, a small number of individuals are randomly chosen from the population and the individual with the highest fitness value is selected as the winner. In roulette wheel selection, each individual is assigned a probability of being selected proportional to its fitness value, and an individual is chosen by spinning a roulette wheel with sections corresponding to each individual's probability. In rank-based selection, individuals are ranked based on their fitness values and a certain proportion of the highest-ranked individuals are selected for reproduction.

Read more

What is an intelligent personal assistant?

An Intelligent Personal Assistant (IPA), also known as a Virtual Assistant or AI Assistant, is a software application designed to assist users with various tasks, typically by providing information using natural language processing. These tasks, traditionally performed by human personal assistants, include reading text or email messages aloud, looking up phone numbers, scheduling, placing phone calls, and reminding the user about appointments.

Read more

It's time to build

Collaborate with your team on reliable Generative AI features.
Want expert guidance? Book a 1:1 onboarding session from your dashboard.

Start for free