New research shows poetic prompts can bypass safety systems in several AI models, revealing vulnerabilities in current guardrail approaches.
New research shows poetic prompts can bypass safety systems in several AI models, revealing vulnerabilities in current guardrail approaches.