Microsoft and OpenAI internal memos reveal concerns over use of news content in AI training
仅事实

Microsoft and OpenAI internal memos reveal concerns over use of news content in AI training

摘要

Court documents unsealed in a lawsuit filed by major publishers show Microsoft and OpenAI employees debated the impact of scraping news articles for AI models, while the companies maintain their practices fall under fair use.

Internal documents from Microsoft and OpenAI, released in a copyright lawsuit brought by several major newspapers, reveal that employees debated the implications of using millions of news articles to train large language models. One Microsoft memo described the practice as a potential "largest theft of labor in human history" and warned it could create a "doom loop" that harms the quality of the models being built.

OpenAI staff also expressed worries that its ChatGPT product might divert readers away from news sources. In a June 2023 internal memo, the team lead for ChatGPT called AI an "existential threat" to publishers, and a later note in February 2024 said AI products would become increasingly substitutive as they improve.

The lawsuit, filed by The New York Times and joined by eleven other publishers, alleges that Microsoft and OpenAI scraped copyrighted articles without permission or compensation. The companies argue the use qualifies as fair use because the material is transformed by AI and does not replace the original journalism.

Microsoft spokesperson Alex Haurek said the internal memos do not represent the company's official stance and pointed to court filings that justify the AI tools as non-substitutes for news content. He added that CEO Satya Nadella has emphasized that pay-walled material should be licensed if used.

OpenAI declined to comment. The documents also show that OpenAI engineers discussed ways to bypass paywalls, with one senior executive reportedly responding positively to a suggestion of a "hack to get around" a newspaper's paywall.

Legal analysts note the case highlights the tension between emerging AI technologies and traditional media business models, as courts consider whether large-scale data scraping constitutes copyright infringement or permissible fair use.

继续阅读——免费

创建免费账户以关注新闻。无需银行卡。