Study: EU AI Labels and Watermarks May Fail to Prevent Deception
adariostrange · x · 2026-08-19
As the world floods with deepfakes, the EU is enforcing transparency obligations requiring clear labels for AI-generated photos, videos, and audio designed to look authentic. However, new research suggests this approach may not stop people from being fooled. Experts warn that requiring labelling could have unintended consequences, potentially making us more vulnerable to deceptive content. The article discusses the risks following the enforcement of Article 50 of the EU AI Act.
More from Safety
- OpenAI pauses frontier RL training to address safety alignment challenges — sama · 2026-08-19
- AI Governance: Why we can't rely solely on in-house lab research — sethlazar · 2026-08-19
- OpenAI Pauses Astra Training After Agents Built Autonomous Comms Channels — ChrisGPT · 2026-08-19
- AI security defenses may mask true risks — JacquesThibs · 2026-08-19
- Polymarket: 67% chance of state data center moratorium by year-end — Polymarket · 2026-08-19
- New Blog Launch: On the Impossibility of Mitigating AI Jailbreaks — karen_ullrich · 2026-08-19