OpenAI and Microsoft knew they were starting a ‘doom loop’ for the web

THE NEWS

What happened

Newly unsealed court documents reveal that OpenAI and Microsoft were fully aware their AI training strategies constituted a 'doom loop' for the web. Internal memos describe their data scraping as the 'largest theft of labor in human history' and admit that chatbots are destroying the economic foundations of publishers like The New York Times, effectively replacing search and driving referral traffic down by up to 60 percent.

CONTEXT

Why it matters

Newly unsealed court documents expose that OpenAI and Microsoft knew their AI training strategies were starting a 'doom loop' for the web. Internal memos describe scraping data as the 'largest theft of labor in human history' and admit chatbots are destroying publishers like The New York Times by replacing search traffic and driving referral traffic down by up to 60 percent.

AT A GLANCE

Key facts

• Internal Microsoft documents explicitly warn that their AI content strategy started a 'doom loop' harming both model performance and the entire web.
• Microsoft's Director of Applied Science, Brent Hecht, called data harvesting the 'largest theft of labor in human history' and admitted it mocks the concept of fair use.
• OpenAI employees acknowledged GPT-4 is 'insanely good at regurgitation' and memorized copyrighted material verbatim despite knowing prevention was important.
• The companies admit their AI summaries are substituting for direct source visits, with search referrals dropping as much as 60 percent due to AI Overviews.
• Executives like Satya Nadella admitted chatbots have replaced the need to go straight to the source for information.