Breaking

NYT Seeks Sanctions, Alleging OpenAI Concealed Evidence in ChatGPT Case

July 9, 2026 at 15:12 EDT

Policy & Regulation
Foundation Models

The New York Times and The Daily News on July 9, 2026, asked the federal court in the Southern District of New York to sanction OpenAI, alleging the company concealed and manipulated evidence during discovery in the ChatGPT copyright lawsuit. The plaintiffs argue that OpenAI's earlier claims of having "no search capability" and facing "privacy concerns" that made compliance difficult were in fact false. According to the report, the filing was also detailed in the NYT's own coverage.

July 9, 2026 · S.D.N.Y. Federal Court

Publishers Ask Court to Sanction OpenAI Over "Hidden" Evidence

The New York Times and The Daily News allege OpenAI concealed and manipulated discovery in the ChatGPT copyright case — claiming its earlier stance of "no search capability" was false, and that billions of output logs were deleted after suit.

78M

De-identified ChatGPT conversations in OpenAI's internal database

Billions

Output logs allegedly deleted or replaced after suit was filed

Dec 2023

Lawsuit begins; main claims allowed to proceed March 2025

The Discovery Sample Dispute

Plaintiffs requested 120 million items. OpenAI narrowed it to 20 million — and the court reportedly deemed the redacted submission "unusable."

120M

Requested

20M

Provided (unusable)

What the deposition allegedly revealed

Searched all along

OpenAI evaluated its training corpus and chat logs internally — even before the suit.

→

"Project Giraffe"

A "Bloom" filter built to detect and log regurgitation of source material in outputs.

→

Logs deleted

Billions of output logs allegedly removed — possibly breaching the preservation order.

Sanctions Sought by Plaintiffs

Bar on evidence

Prohibit use of the 20M-item sample

Finding of fact

Rule that logs showed regurgitation of content

Limits on rebuttal

Restrict OpenAI's counterarguments

Cost shifting

Require OpenAI to cover added legal fees

Publishers' position

Millions of articles were used in training without permission, and ChatGPT reproduced them near-verbatim or fabricated articles under the NYT's name — harming subscription and ad revenue.

OpenAI's defense

Flatly denies the allegations, calls the effort a privacy intrusion as the case weakens, cites protection of user conversations, and maintains its fair use / transformative-output defense.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.