Microsoft and OpenAI feared AI would starve the news it feeds on

• Microsoft’s director of applied science called the mass scraping of news “the largest theft of labour in human history”.¹
• OpenAI’s head of ChatGPT warned that products trained on publishers’ work posed an “existential threat” and were “largely substitutive”.¹
• Microsoft data cited in court showed click-through rates to some news sites falling by up to 93% when users received AI-generated answers instead of conventional search results.¹
• The filings describe datasets containing at least 160,903 unique works from news publishers, including more than 91,692 copies in OpenAI mid-training datasets.²
• The evidence comes mainly from the publishers’ legal brief; many underlying exhibits remain sealed, and the judge has not yet ruled on its significance.²

The most damaging revelation is not that Microsoft and OpenAI understood AI could disrupt journalism. It is that their own internal messages appear to describe the central business risk in unusually direct terms: the products could destroy the economic supply chain that produces the information they need.

Microsoft director Brent Hecht wrote that an AI product threatening its suppliers was creating a “doom loop”. OpenAI executive Nick Turley reportedly said there was “no good reason to click” when a chatbot supplied the answer itself. Those statements matter because the copyright dispute turns partly on whether the companies’ use of publishers’ work transformed it into something new, or helped build products that substitute for the original reporting.¹

The filings also allege that OpenAI and Microsoft moved news content between projects originally associated with search and AI training, assembled large publisher datasets, and sought ways around paywalls. Those are allegations by the news organisations, not findings by the court. The companies’ responses have also stressed that individual employees’ views were not legal conclusions or official company positions.¹

But the commercial consequence is already clear enough to understand the stakes. If AI systems answer questions without sending readers to the reporting underneath, publishers lose the visits, subscriptions and advertising that finance the next story. The companies then risk weakening the very reporting their systems rely on — while arguing in court that taking it without a licence is lawful.²

Footnotes

1. Ars Technica, “Microsoft exec called AI scraping the ‘largest theft of labor in human history’” ↩ ↩² ↩³ ↩⁴ ↩⁵

2. TechCrunch, “Microsoft exec called AI scraping ‘the largest theft of labor in human history,’ new unredacted filings reveal” ↩ ↩² ↩³