Make watermarking mandatory for all AI-generated text. Transparency should outweigh concerns about detection limitations and personal preferences.

Require watermarking only in high-risk settings. Political communication, academic submissions and official information should receive special scrutiny.

Let users decide whether to enable watermarking. Individuals should retain control over how AI assistance is disclosed.

Delay wider adoption until independent tests establish greater reliability. False accusations and missed AI content could undermine public trust.

Reject invisible watermarking as a standard approach. Undetectable signals controlled by technology companies raise concerns about transparency and digital autonomy.