Palo Alto Networks, Friday, August 28th, 2026
Perturbation Probing: A New Diagnostic for the Fragility of LLM Safety
Unit 42 research finds LLM safety refusal behavior lives in a thin neural layer, arguing for external security controls.
more →
2 articles that week
Palo Alto Networks, Friday, August 28th, 2026
Unit 42 research finds LLM safety refusal behavior lives in a thin neural layer, arguing for external security controls.
more →
Palo Alto Networks, Tuesday, August 25th, 2026
Unit 42 surveys AI-enabled malware and argues existing behavioral detection stops AI-authored code before execution.
more →