OpenAI, Microsoft knew AI posed ‘existential threat’ to news industry, filings reveal

# Internal OpenAI and Microsoft Discussions Highlight AI Threat to News Publishers

Internal communications from OpenAI and Microsoft show that some employees had recognized the potential impact of generative artificial intelligence on the news industry even before The New York Times filed its copyright lawsuit against the companies.

The disclosures are part of a recent court filing in the ongoing legal battle between The New York Times, OpenAI and Microsoft. The newspaper alleges that the companies used millions of copyrighted articles to develop AI systems without authorization or payment, while their products could potentially provide users with information without requiring them to visit the original publishers’ websites.

According to the filing, Nick Turley, an OpenAI executive involved with ChatGPT, warned internally that publishers faced what he described as an “existential threat” from AI products. He reportedly characterized the products as already substantially substituting for some publisher content and suggested that this effect could increase as AI systems became more capable.

Internal comments attributed to Microsoft Director of Applied Science Brent Hecht also raised concerns about the relationship between AI companies and the publishers whose content contributes to the information ecosystem. Hecht reportedly argued that large-scale copying of copyrighted material could create serious economic risks for the organizations producing that content.

Hecht's concerns centered partly on the possibility that AI-generated answers could reduce the number of users clicking through to original news websites. Lower traffic could mean reduced advertising and subscription revenue for publishers, potentially affecting their ability to finance journalism and generate the fresh information that AI companies themselves rely on.

The court filing also points to internal discussions surrounding access to paywalled material. According to the filing, OpenAI President Greg Brockman responded positively after a researcher reported finding a method to get around The New York Times’ paywall. The disclosure has become another point of contention in the copyright dispute.

The filings further cite testimony from Microsoft CEO Satya Nadella. According to the Financial Times' account of the documents, Nadella said that had he known OpenAI had trained models on material behind a paywall, he would have exercised Microsoft's contractual rights to require the models to be retrained. Microsoft has said that Nadella’s testimony addressed broader principles and did not represent a conclusion on the copyright questions at issue.

Microsoft has also sought to distance itself from Hecht’s comments, saying they represented an individual employee’s perspective rather than the company’s official position or a legal analysis.

The disclosures highlight a central tension in the copyright litigation: AI companies argue that using publicly available material to train large language models can qualify as fair use, while publishers contend that the same practices can create commercial products that compete with the original sources.

OpenAI has previously argued that its models transform the material used during training rather than simply reproducing publishers’ articles. The company has also disputed the characterization that its products function as direct substitutes for news organizations.

The New York Times filed its lawsuit against OpenAI and Microsoft in December 2023, making it one of the most prominent copyright cases involving generative AI. Other publishers and newspaper groups have subsequently brought similar claims, contributing to a broader wave of litigation over the use of copyrighted material to train AI systems.

The dispute reflects a larger question facing the technology and media industries: how should copyright law apply when AI developers use enormous collections of existing content to build systems capable of generating new responses and information?

For news organizations, the concern extends beyond the initial use of their articles. Publishers have increasingly focused on whether AI-generated answers could divert audiences away from original reporting, potentially weakening the traffic, subscriptions and advertising revenue that support journalism.

For AI companies, the cases could help establish how copyright law applies to the training of large language models and what permissions, licensing arrangements or other restrictions may be required when copyrighted material is used at scale.

The latest disclosures therefore add another dimension to the legal dispute, showing that at least some employees within the companies had discussed the potential economic consequences for publishers while the technology was still developing. The court case will ultimately turn on the legal arguments and evidence presented by the parties rather than the concerns expressed in individual internal communications.