Microsoft Executive Calls AI Training Data Use ‘Largest Theft of Labour’

A Microsoft executive has described the use of internet content to train artificial intelligence models as an “astonishing theft,” calling it potentially the largest theft of labour in human history. The remarks by Brent Hecht, Microsoft’s director of applied science, emerged in unsealed court filings from the copyright dispute involving The New York Times, Microsoft and OpenAI.

Hecht wrote that millions of people could view AI models collecting their work without permission or compensation as an unprecedented form of theft. His comments highlighted concerns over creators’ content being used to develop AI systems despite many creators not intending for their work to be used in this way.

Microsoft and OpenAI have defended their use of copyrighted material by arguing that AI training falls under fair use and transforms the material rather than merely reproducing it. The New York Times and other publishers dispute this, arguing that AI products can compete directly with the original sources.

Advertisement

The court filings also point to concerns about AI chatbots reducing traffic to news websites. Microsoft data cited in the documents showed click-through rates for New York Times and Daily News domains were 83% to 93% lower through Copilot’s answer engine than through traditional Bing Search. OpenAI executives also discussed AI products becoming increasingly substitutive of original sources.

The New York Times filed its lawsuit against Microsoft and OpenAI in 2023, with other news organisations later joining the copyright dispute. The companies continue to rely on their fair-use defence, while the publishers are seeking a ruling in their favour.