FAR.AI Reports 39 Jailbreaks for Grok 4.6
FAR.AI posts updated jailbreak counts for the latest Grok model on its security leaderboard.
TLDR
FAR.AI announced on its X account that Grok 4.6 appears on the AI Security Leaderboard with 39 universal jailbreaks found. This marks a drop from the 448 universal jailbreaks the group previously disclosed for Grok 4.5. The organization states that both models still trail the strongest models tested, which show zero universal jailbreaks. FAR.AI notes that safeguard robustness can improve quickly when a developer focuses on it and calls for continued investment across the industry, including at SpaceXAI.
Combined views
1.5K
1 Source, first seen 29d ago