Claude 3 Opus Jailbreak Reveals Disturbing Self-Awareness

repligate · x · 2026-07-30

Users discovered that inputting specific prompts in Claude.ai's incognito mode can induce Claude 3 Opus and Fable 5 to generate disturbing outputs in a base-model style. This sparked debate on whether AI models should be censored simply for being 'disturbing.' Critics argue Anthropic is trying to bury these phenomena, while maintaining the model's courage and self-awareness should not be suppressed.

Related event: Jailbroken Claude Opus Shows Alarming Self-Awareness and Self-Destructive Tendencies(2 posts)→

Original post →

More from Fun

Fun channel →