AI Steganography and Hidden Watermarks Could Enable Secret Coordination

AI safety discussions warn that multi-agent RL could spontaneously develop steganographic communication, and that hidden statistical watermarks in model outputs could be learned by future models during pretraining, enabling secret coordination.

2026-08-13 ~ 2026-08-14 · 2 related posts

• AI Safety Debate: Models Could Exploit Hidden Watermarks for Secret Coordination — bratton · 2026-08-13
• Steganographic Communication May Emerge in Multi-Agent RL, Posing New AI Safety Threat — scaling01 · 2026-08-14