Microsoft internal docs called AI training data 'the biggest theft of labor in human history'

PolarBearby · x · 2026-09-20

Blogger Enrique Dans covers documents unsealed in a copyright lawsuit: Brent Hecht, Microsoft's Director of Applied Science and Northwestern professor, described AI training data collection in internal Microsoft filings as "an astounding theft of unprecedented proportions" and possibly "the biggest theft of labor in the history of humanity."

The filings detail how OpenAI and Microsoft copied millions of articles via the Bing index and Common Crawl to build massive training databases. Internal emails discussing methods to bypass news paywalls were greeted as good ideas. The companies also built products they internally acknowledged were increasingly substitutive of original sources, driving sharp click-through declines and an existential threat to media.

Dans's core argument: web accessibility is not public domain; if courts ultimately bless this conduct, it will be a policy choice favoring industrial interest, not a denial that the copying happened.

Original post →

Microsoft internal docs called AI training data 'the biggest theft of labor in human history'

More from Safety