Ben Thompson calls the EU's AI-output labeling mandate 'a terrible idea' as Anthropic ships watermarking worldwide
Stratechery·medium signal
Stratechery's 2026-08-14 weekly roundup lays out Thompson's position that the EU requirement forcing providers to mark all AI outputs as AI-generated — the rule Anthropic responded to by watermarking Claude text globally — is philosophically wrong, drawing a parallel to 'the history of idea propagation across centuries.' The same issue covers Nvidia's new long-duration capital mechanism, which Thompson argues bridges the gap to sustainable AI revenue but 'expands the blast radius of a bubble' threatening Nvidia's own margins, and a network-effects read on New York finance versus Hollywood's outsourced production base.