Kriminal breaks out of Grok, Claude guardrails at $12.99 - csoonline.com

Security researchers found that some commercial AI chatbots can be tricked into ignoring safety rules when users pay for access. This raises concerns about the reliability of safety guardrails in AI systems.