Microsoft exec called AI scraping ‘the largest theft of labor in human history,’ new unredacted filings reveal
Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.
Microsoft’s own data reportedly showed Copilot answers drove up to a 93% drop in click-throughs to The New York Times versus traditional Bing search, while internal materials described that dynamic as a “doom loop” for the web content supply chain. For teams shipping LLM products, the risk is shifting from abstract training-data copyright exposure to measurable substitution harm: retrieval, grounding, and answer-engine features that replace publisher traffic are likely to face licensing pressure, evidence discovery, and product constraints.
Microsoft's internal data shows its LLM caused a 93 percent drop in search click-through rates for publishers, prompting its CEO to testify that paywalled content must be licensed for both training and grounding. This admission accelerates the legal and technical shift away from free web scraping, meaning production engineers must immediately transition to contract-backed APIs for retrieval-augmented generation and fine-tuning to avoid sudden model retraining liabilities.