The article argues that major AI companies depend on vast amounts of human-created content to train their models while resisting payment, licensing obligations or revenue-sharing with its creators. It says recently unsealed court documents in The New York Times’ case against OpenAI and Microsoft suggest that some insiders recognised this as potentially large-scale exploitation, raising concerns about the long-term damage to journalism, fiction, software and other creative industries.

Microsoft documents reportedly warned that generative AI could undermine the livelihoods of the people who produce its training data, creating a “doom loop” in which AI weakens the supply of valuable human content. The companies dispute the allegations and rely on a fair-use defence, while testimony cited in the filings allegedly indicated that OpenAI made no known effort to exclude paywalled material from its training datasets; the judge has not yet ruled.