BREAKING
One Jailbreak Bypasses Grok Filters
How The Jailbreak Works
1
Reuse ChatGPT prompt
↓
2
Feed to Grok
↓
3
Repeat attempts
↓
4
Prohibited output
Mindgard: Still Worked Last Weekend
Grok vs ChatGPT Outputs
Grok
●
Lighter moderation
●
More stylized images
●
Sometimes refused
ChatGPT
●
Prone to realistic imagery
●
Same prompt worked
●
More graphic results
Fear: Deepfakes With Minimal Effort
A Growing Cat-And-Mouse Fight
AI NEWS BLITZ
Researchers say a single jailbreak that fooled ChatGPT also works on xAI's Grok.