AI Summary
5 min readIn February 2025, Anthropic announced it would begin embedding a secret, invisible signal—a cryptographic watermark—into every piece of text generated by its AI models, including Claude. The move was made to comply with a new European Union transparency code, part of the broader EU AI Act. But Anthropic chose to apply the watermark worldwide, not just in Europe. That means every time a user in the United States asks Claude to draft an email, review a contract, or write code, the output carries a hidden marker designed to travel with the text even after it is copied, pasted, and reposted. Neil Chilson, a former chief technologist at the FTC and now head of AI policy at the Abundance Institute, joined Taylor Lorenz to explain how this watermarking actually works, why it comes with serious trade-offs, and why the policy may end up creating more problems than it solves.
How Text Watermarking Works
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of Taylor Lorenz’s Power User
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 (00:01) **Episode Introduction & Ad Reads** - Host Taylor Lorenz sets up the topic: Anthropic will watermark all AI-generated content to comply with an EU law, and she argues this creates serious privacy and speech problems.
- 2 (04:43) **What Is Watermarking?** - Neil Chilson defines a digital watermark as a hidden mark embedded in content to prove its origin, distinct from metadata.
- 3 (06:36) **The EU AI Act's Transparency Code** - The EU's 2021 law, with provisions taking effect in 2024 and 2025, mandates that generative AI providers watermark their outputs.
- 4 (08:20) **How Anthropic's Text Watermark Likely Works** - Chilson describes the probable algorithm: the AI selects from plausible word choices (like "mango" vs. "lychee") in a biased way that a secret key can later detect.
- 5 (11:31) **Initial Problems with the Watermark** - Lorenz asks what issues stand out about the policy, and Chilson points out that the watermark adds an unintended "message" to the user's prompt.
- 6 (14:41) **Precision Problems: Legal, Scientific, and Creative Text** - Lorenz raises concerns about using AI for hyper-specific tasks like legal documents, scientific research, and poetry, where word choice is critical.
- 7 (18:24) **Cybersecurity and Privacy Vulnerabilities** - Lorenz asks about security risks from creating machine-readable watermarking systems.
+ Full timestamped outline available in the app
Guests on this episode
Show Notes
To access an ad-free listening experience, get bonus episodes, and receive my weekly newsletter, subscribe to my Patreon or Substack 👇🏻
Buy a paid subscription to my newsletter at https://www.usermag.co
Support my work on Patreon: http://patreon.com/taylorlorenz
Anthropic just announced that Claude and all its AI tools will now watermark every piece of text a user generates, even if it's scientific research or computer code. In this episode of Free Speech Friday, I sit down with Neil Chilson, former FTC Chief Technologist, to break down exactly what AI watermarking is, how the EU AI Act transparency code is forcing this change worldwide, and why this could threaten your privacy and free speech.
We explain how probabilistic text watermarking works by subtly altering word choices in AI-generated content, why this creates serious problems for legal documents scientific research and creative writing, and how Anthropic's decision to apply these watermarks globally affects US users who never consented to this tracking.
Neil also breaks down the cybersecurity risks, the potential for government overreach, and why open-weight AI models might be the only escape hatch from extraterritorial regulation.
If you use Claude, ChatGPT, Gemini, or any AI writing tool, (or even if you don't!) you need to understand how the EU AI Act transparency requirements are reshaping AI output behind the scenes. We discuss the dangers of compelled speech on digital platforms and what watermarking means for the future of anonymous online expression.
SUPPORT MY WORK:
Buy a paid subscription to my newsletter at https://www.usermag.co
Support my work on Patreon: http://patreon.com/taylorlorenz
More from this podcast
Taylor Lorenz’s Power User →