The demand for conversations related to AI has skyrocketed, with a staggering request of 1.4 billion interactions originally submitted by The New York Times. This figure has since been negotiated down to 20 million randomly sampled ChatGPT conversations, pushing OpenAI into a complex legal battle. Security experts believe that this case could significantly alter how data retention practices are approached across the AI sector.
OpenAI’s Legal Challenges
OpenAI received a court-preservation order on May 13, which mandates the company to retain all output log data that would typically be wiped away, regardless of user deletion requests or existing privacy regulations. District Judge Sidney Stein upheld this decision on June 26 after an appeal from OpenAI, dismissing claims that user privacy should take precedence over litigation needs.
Clash of Privacy and Legal Commitments
This preservation order forces OpenAI to keep data from ChatGPT and API users indefinitely, which conflicts with the company’s usual practice of deleting conversations after 30 days if users haven’t opted to retain them. The requirements apply to user data collected from December 2022 through November 2024 and impact subscribers across various tiers, including ChatGPT Free, Plus, Pro, and Team members, as well as API customers who lack Zero Data Retention agreements.
However, users under ChatGPT Enterprise, ChatGPT Edu, and those with Zero Data Retention contracts are not subject to these preservation mandates. The order does not change OpenAI’s policy of not using business data for model training by default.
To manage this sensitive data, OpenAI has instituted restricted access protocols, ensuring that only a small, audited team focused on legal matters can access preserved data. The company emphasizes that this information will not be shared with The New York Times, the court, or any external entities at this stage.
The Copyright Lawsuit Explains Data Preservation Needs
The New York Times launched its copyright infringement lawsuit in December 2023, arguing that OpenAI unlawfully utilized millions of the newspaper’s articles to train its language models, including ChatGPT and GPT-4. The lawsuit alleges that this action constitutes both copyright infringement and unfair competition, contending that OpenAI reaps profits from protected material without authorization or compensation.
Beyond seeking financial restitution, The Times is asking for the destruction of all GPT models that utilized its copyrighted works, which could potentially lead to damages running into billions of dollars, both in actual and statutory claims.
In making its case for preservation, The Times illustrated a precedent where another AI company willingly handed over 5 million private user chats in an unrelated legal matter. OpenAI, however, dismissed this comparison as not applicable to their situation.
Technical and Regulatory Challenges Ahead
Meeting the requirements for indefinite data retention also poses various technical hurdles. OpenAI needs to develop robust systems for storing vast amounts of user conversations from around the globe, which involves considerable time and financial resources.
Moreover, the preservation order presents challenges concerning international data protection laws like GDPR. Although OpenAI’s terms of service stipulate data preservation for legal obligations, the company argues that the demands from The Times exceed reasonable limits and disregard established privacy standards.
To address these issues, OpenAI proposed several privacy-conscious alternatives, such as performing targeted searches among the preserved samples to locate conversations that may reference New York Times articles. These alternatives aim to deliver relevant data while protecting broader privacy concerns.
Recent court rulings have provided some relief, allowing OpenAI to stop preserving all new chat logs from September 26, 2025, moving forward. Nevertheless, the company is still required to retain previously saved data and must hold onto information from ChatGPT accounts specified by The New York Times.
Implications for AI Governance
This legal case brings to the forefront immediate challenges concerning privacy for the 400 million users of ChatGPT. Security professionals warn that the preservation order disrupts long-standing assumptions surrounding data deletion in AI interactions.
OpenAI’s CEO, Sam Altman, suggested that ongoing discussions about “AI privilege” are essential, implying that interactions with AI should enjoy protections akin to attorney-client privilege. The idea is to frame unlimited data storage as a precarious step regarding privacy in AI communications.
The ramifications of this litigation could be significant for enterprise clients pairing ChatGPT with sensitive applications. Organizations that utilize OpenAI for sectors like healthcare, legal services, or finance may need to reassess their compliance with regulations such as HIPAA and GDPR in light of the new data preservation obligations.
Legal experts caution that the outcome of this case could pave the way for third-party discovery, complicating how preserved AI conversation logs might be accessed in unrelated legal disputes. This not only presents privacy concerns but could also impinge on attorney-client privilege protections.
Ultimately, the implications extend beyond immediate legalities, influencing how AI firms gather and utilize training data, which could reshape future AI development and deployment. Crucial questions remain unresolved regarding how the fair use doctrine applies to AI model training and the boundaries of discovery in AI-related copyright cases.
