Anthropic’s Best-of-N jailbreak technique proves how introducing random characters in a prompt is often enough to successfully bypass AI restrictions.