July 9, 2026, (Inside AI) — A coalition of newspapers led by The New York Times asked a federal court in Manhattan on Thursday to sanction OpenAI, alleging the AI company lied about its ability to search its systems for copyrighted content used in training ChatGPT.
The motion, filed in the U.S. District Court for the Southern District of New York, claims OpenAI falsely stated it could not search its large language models for the newspapers’ articles while secretly having done so before the first lawsuit was even filed. The newspapers also accuse OpenAI of deleting or rendering unsearchable billions of relevant ChatGPT conversations.
The plaintiffs are seeking sanctions including attorneys’ fees and a court finding that OpenAI’s chat logs prove misuse of copyrighted works. OpenAI spokespeople did not immediately respond to a request for comment.
The original lawsuit, initiated by The New York Times in 2023, accuses OpenAI and its largest backer Microsoft of using millions of articles without permission to train the models behind ChatGPT. The case is a bellwether in a wave of copyright disputes against AI developers, including Anthropic and Meta, brought by authors, artists, and music labels.
The Deletion Allegation and Its Broader Implications
The filing reveals a deeper rift: OpenAI allegedly deleted billions of ChatGPT conversation logs, which the newspapers argue could have contained evidence of infringing outputs. This accusation moves beyond simple discovery disputes into potential spoliation of evidence, a serious charge that could reshape how AI companies manage training data records.
Legal experts note that if proven, such deletion could trigger adverse inference rulings—where a court assumes the destroyed evidence was unfavorable. This would be unprecedented in AI copyright litigation and could force companies to implement rigorous data retention policies, fundamentally altering AI development practices.
The newspapers’ lead attorney, Ian Crosby, stated:
"For over two years, OpenAI lied to The Times, The Daily News Plaintiffs, the public, and the court. It claimed searching ChatGPT outputs for copies of The Times' and the Daily News Plaintiffs' content was infeasible, burdensome, and invasive of users' privacy - while at the same time concealing that it had already done such searches."
This statement underscores the core allegation: OpenAI’s public and legal posture contradicted its internal capabilities, potentially misleading the court about the feasibility of detecting copyright infringement in AI outputs.
Technical Feasibility and the Privacy Paradox
At the heart of the dispute is a technical question: Can AI models be reliably searched for specific copyrighted content? OpenAI had argued that searching training data or output logs was impractical due to the scale and structure of models like GPT-4. However, the newspapers claim an OpenAI employee testified that the company had already performed multiple searches for their content, suggesting such searches are not only possible but were conducted covertly.
This revelation challenges the industry’s common defense that AI systems are “black boxes” incapable of tracing individual data points. Researchers have demonstrated methods for membership inference attacks that can determine if specific data was used in training, but these techniques are often probabilistic. The newspapers’ filing implies OpenAI may have more deterministic search capabilities than previously disclosed.
The privacy argument adds another layer. OpenAI claimed that searching user conversations would violate privacy, yet the newspapers argue this was a pretext. If OpenAI did search chats without user consent, it could face separate privacy scrutiny, creating a Catch-22 for the company: either it lied about the searches or it violated user privacy.
The case is part of a broader reckoning over fair use in AI training. While tech companies argue that training on publicly available data is transformative, publishers contend it undermines their business models. The outcome here could set a precedent for dozens of pending cases, potentially requiring AI developers to license training data or face stricter discovery obligations.
As the litigation unfolds, the focus will likely shift to the forensic examination of OpenAI’s data practices. The court’s response to the sanctions motion may signal how seriously it views the alleged deception, with possible repercussions for AI transparency standards industry-wide.
