Tom Chavez, co-founder of super{set} and former CEO of Rapt, argued on LinkedIn this week that handing proprietary data to large language models repeats the bargain publishers struck with Google two decades ago, and that no reliable technique exists to reverse it. The post, reviewed on Saturday, August 8, 2026, showed 80 reactions, two comments and three reposts. Chavez wrote, "20 years ago, I walked around Manhattan begging digital publishers not to hand their ad inventory to Google for pennies on the dollar." His LinkedIn profile lists him as chief executive and co-founder of Rapt from 1999 to 2008, a San Francisco company acquired by Microsoft, followed by roles as general manager for online publisher businesses at Microsoft from April 2008 to October 2009 and entrepreneur in residence at Accel Partners from October 2009 to May 2010. He organizes internet commerce into three phases: a pre-Facebook era where he urged publishers "to forego the nickel today and invest in the ability to sell their advertising for dollars tomorrow," a corrective phase where companies learned "Data is the coin of the realm; don't cede it to the giants," and the present phase where LLMs, "worth trillions," pitch companies to give up enterprise data with the assurance "Your data still belongs to you." Chavez added, "It reminds me of Google's pitch in the early 2000's." The post rests on an assertion about machine