OpenAI and Microsoft knew they were starting a ‘doom loop’ for the web
What happened
Court documents unsealed in the New York Times lawsuit against OpenAI and Microsoft reveal internal warnings from the companies about the risks of their AI training methods. The documents describe the scraping of web data to train AI models as the “largest theft of labor in human history” and say it created a “doom loop” that harms the web ecosystem. Microsoft’s Director of Applied Science, Brent Hecht, provided many of these candid assessments.
Why it matters
The internal language shows OpenAI and Microsoft recognized early that their approach to data scraping would degrade web content quality and push the internet into a negative feedback loop. This “doom loop” means web publishers could face losses in traffic and revenue as AI systems mine content without permission or compensation, undermining the very data sources that fuel AI improvements. The firms’ own framing of this as a mockery of fair use challenges prevailing legal and ethical assumptions about training data for AI.
For builders and operators, this raises questions about the sustainability of using scraped data at scale to train models. It pressures AI companies to rethink data sourcing strategies and exposes legal and reputational risks that could slow down large-scale AI training efforts relying on unlicensed web content.
What to watch next
Regulators and courts are likely to scrutinize AI training practices more aggressively, with this lawsuit setting a precedent for how data use is treated. Businesses licensing or creating AI models should monitor the evolving legal landscape around fair use and data rights to avoid liability. OpenAI and Microsoft may also need to innovate alternatives to unfettered web scraping, such as partnering with content creators or investing in proprietary datasets. The “doom loop” concept may drive companies to balance model performance with respecting content ownership to ensure health of the web ecosystem over the long term.
AI Quick Briefs Editorial Desk