NYT accuses OpenAI of bypassing paywall

The New York Times accuses OpenAI of deliberately and misleadingly circumventing the newspaper’s paywall to collect content for AI training. Newly released court documents show internal messages revealing that an OpenAI researcher shared a trick to bypass the paywall, to which then-president Greg Brockman responded positively. At the same time, the documents contain striking internal warnings about the consequences of AI for journalism.

In brief:

• Court documents show internal messages in which OpenAI employees deliberately circumvented the NYT paywall.
• A Microsoft executive called AI training “the greatest theft of labour in human history”.
• OpenAI’s ChatGPT chief admitted internally that AI products largely replace journalistic content.

Internal messages about circumventing the paywall

The released court filings show that OpenAI researcher Nick Ryder informed then-president Greg Brockman about “a trick to bypass the NYT paywall” during conversations about scraping the website. According to the documents, Brockman responded with “ah nice.”

The New York Times took OpenAI to court over its use of its journalistic content to train AI models. The now-released internal communication plays a central role in that lawsuit and sheds new light on how deliberately OpenAI acted.

Internal warnings about AI and journalism

The court documents also contain notable statements from employees at both OpenAI and Microsoft about the broader consequences of AI for the media sector. A Microsoft executive described training AI on journalistic content as “the greatest theft of labour in human history”.

Nick Turley, responsible for ChatGPT at OpenAI, admitted internally that AI products “largely replace” journalistic content and that this effect will only grow as the technology improves. That admission stands in sharp contrast to OpenAI’s public communications, which usually emphasise that AI tools are complementary rather than a replacement.

Lawsuit touches on broader debate about AI training

The case between The New York Times and OpenAI is part of a growing number of legal conflicts over the use of copyrighted material for AI training. Publishers and journalists worldwide are wondering whether and how they can demand compensation for the use of their content.

OpenAI frequently finds itself in the firing line when it comes to the security and use of its systems. We previously reported on a hack in which OpenAI was attacked via Anthropic’s AI model. The new court documents now add an internal ethical question about how the company handles other people’s content.