Anthropic's Claude watermark fails to stop cheating
adariostrange · x · 2026-08-20
Anthropic introduced invisible watermarking for Claude's text and file outputs to comply with the EU AI Act. However, experts argue it won't effectively stop academic cheating, as AI lacks understanding and detection mechanisms remain limited.
More from Safety
- Platforms fight AI slop with filters, labels, and restrictions — emmanuelvivier · 2026-08-20
- Users cancel Claude subscriptions over persistent invisible watermark concerns — emmanuelvivier · 2026-08-20
- MPA Strikes Deal with ByteDance to Protect IP in AI Video/Image Models — emmanuelvivier · 2026-08-20
- Max Spero on AI detection bias and writing assistance experience — amplifiedamp · 2026-08-20
- WIRED: Flock's Police AI Can Track People Despite Denials — zacharynado · 2026-08-20
- Fabraix automates red teaming, exploits bank agent with invoice image — aftahi_ai · 2026-08-20