OpenAI responds to The New York Times data preservation request

OpenAI has publicly responded to demands from The New York Times for the indefinite preservation of user data, framing the request as an overreach that jeopardizes user privacy without substantially aiding the ongoing legal proceedings. The company’s stance highlights a critical juncture in the intersection of artificial intelligence development, data privacy, and intellectual property rights.

The core of the dispute lies within a copyright infringement lawsuit initiated by The New York Times against OpenAI and Microsoft. The Times alleges that OpenAI unlawfully used its copyrighted content to train AI models, including ChatGPT, leading to outputs that sometimes replicate or closely mimic Times articles verbatim. This practice, the lawsuit contends, not only infringes on copyright but also threatens the economic viability of journalism by creating substitutive products without permission or payment.

The Genesis of the Data Preservation Dispute

The New York Times’ legal action against OpenAI and Microsoft began in December 2023, with allegations of copyright infringement stemming from the use of news articles to train AI models. As the litigation progressed, The Times sought to compel OpenAI to retain specific user data, including deleted conversations, for an extended period. This demand was based on the speculation that such data might contain evidence supporting their claims.

OpenAI, in response, has vociferously opposed this data preservation request, labeling it as excessive and potentially harmful to user privacy. The company argues that such a sweeping demand could expose sensitive user information and undermine the trust users place in their services. This contention is central to OpenAI’s public defense, as articulated in statements and blog posts.

The legal maneuvering intensified when a magistrate judge ordered OpenAI to preserve all output log data that would otherwise be deleted, irrespective of user deletion requests or privacy regulations. This order, issued in May 2025, was a significant development, compelling OpenAI to retain data that was previously subject to its standard 30-day deletion policy for ChatGPT conversations and API data. OpenAI subsequently appealed this order, emphasizing the conflict with its privacy commitments and the burdens of indefinite data retention.

OpenAI’s Stance on User Privacy and Data Handling

Central to OpenAI’s defense is its commitment to user privacy. The company asserts that its standard data retention practices are designed to protect users, with deleted conversations typically removed from its systems within 30 days. This policy is intended to provide users with assurance that their interactions are not stored indefinitely.

However, the court-ordered preservation requirement created a substantial archive of personal interactions that were not originally intended to persist beyond short-term use. OpenAI has stated its intention to strip identifying details from the records, but privacy experts caution that de-identification offers limited protection, as unique phrasing and contextual clues can still allow for re-identification.

OpenAI’s public pushback against the data preservation demands is also framed as a defense of user trust. By making its stance public, the company aims to foster a broader conversation about the obligations AI companies have to protect user privacy during legal disputes and the extent to which courts can demand access to user interactions. This approach seeks to establish a precedent for prioritizing user trust alongside legal defense.

The company’s argument also touches upon the technical feasibility of such broad data requests, suggesting that fulfilling them can be exceedingly complex and potentially unfeasible given the scale of operations. OpenAI has also introduced the concept of “AI privilege,” a proposed framework that would afford certain AI-facilitated conversations protections similar to those in doctor-patient or attorney-client relationships.

The Legal Framework: Copyright, Fair Use, and Discovery

The underlying lawsuit hinges on copyright infringement claims. The New York Times alleges that OpenAI’s training of AI models on its published articles constitutes unauthorized copying. OpenAI, conversely, argues that its use of publicly available internet materials for AI training falls under the doctrine of fair use, asserting that such use is transformative and serves a different purpose than the original works.

The legal battle highlights the complexities of discovery in the digital age. The Times seeks access to user logs and data as part of its discovery process, aiming to find evidence that supports its infringement claims. OpenAI views these discovery demands as disproportionate and invasive, arguing they are inconsistent with its operational processes and privacy policies.

A key point of contention in copyright law, particularly in AI contexts, is the concept of “substantial similarity.” While copyright holders must typically prove that AI outputs are “substantially similar” to their original works to establish infringement, some plaintiffs argue that the mere use of copyrighted materials for training is sufficient, especially if they can demonstrate direct evidence of their content being ingested by AI models.

OpenAI’s defense strategy has also involved accusations of “prompt hacking,” suggesting that plaintiffs may have manipulated ChatGPT with specific prompts to generate outputs that resemble copyrighted material. This defense aims to portray such outputs as anomalous results, achievable only through extensive efforts, thereby weakening the claim of direct infringement or limiting damages.

Implications for AI Development and User Data

The data preservation dispute and the broader copyright lawsuit have significant implications for the future of AI development. If courts compel broad data retention mandates, it could create substantial privacy risks for users and impose significant operational burdens on AI companies.

Conversely, if OpenAI’s arguments for data minimization and user privacy prevail, it could set a precedent for how AI companies handle user data in the face of legal discovery requests. This would underscore the importance of robust data governance policies and transparent user agreements.

The legal outcomes of these cases could shape the landscape of AI training data, influencing whether companies must seek explicit permission or licensing agreements for all data used in model development. This, in turn, could impact the accessibility and cost of developing advanced AI systems.

Furthermore, the case raises questions about the balance between innovation and intellectual property rights. OpenAI contends that limitations on data usage could stifle the progress of AI, while copyright holders argue that their rights must be protected to ensure the continued creation of quality content.

Evolving Legal Landscape and Future Considerations

The legal battles surrounding AI training data and user privacy are ongoing, with various court rulings providing evolving interpretations of copyright law and data protection principles. Some rulings have favored AI developers by dismissing claims based on a lack of demonstrated harm or by affirming fair use. Other decisions have compelled AI companies to preserve data, highlighting the courts’ role in ensuring evidence is available for litigation.

The outcome of OpenAI’s response to The New York Times’ data preservation request will likely have far-reaching consequences. It could influence regulatory approaches to AI data usage, user privacy expectations, and the legal frameworks governing the development and deployment of artificial intelligence technologies worldwide.

As the legal landscape continues to shift, businesses and individuals alike must stay informed about these developments. Understanding the nuances of AI data governance, copyright law, and privacy regulations is becoming increasingly crucial in navigating this complex and rapidly evolving technological frontier.

The ongoing litigation underscores a critical challenge: how to foster innovation in AI while upholding fundamental rights to privacy and intellectual property. The decisions made in cases like this will shape the ethical and legal boundaries of artificial intelligence for years to come.