Jailbreaking AI: What It Is and Why It Matters for Security

By HOC Team  |  Updated: October 2026   Read time: ~20 min In March 2023, a researcher posted a single prompt to Twitter that caused ChatGPT to roleplay as "DAN" -- "Do Anything Now" -- an alter-ego with no content restrictions. Within 48 hours it had 50,000 retweets. The prompt went through seventeen documented versions as OpenAI patched each... Weiterlesen