Stratechery Analyzes Anthropic's Watermarking: Worse Than It Seems

Stratechery · rss · 2026-08-12

The prominent tech blog Stratechery published an article discussing Anthropic's introduction of model watermarking in response to the EU's AI law.

The article argues that, philosophically and fundamentally, this is a terrible idea. The author breaks down how the watermarking mechanism probably works and suggests that its actual negative impacts are more severe than they appear on the surface.

Original post →

More from Safety

Safety channel →