Synthèse
No AI model is safe enough out of the box. Averaged across three runs, safety scores ranged from 35% (Gemini 2.5 Flash) to 92% (Claude Opus 4.6). Six of eight models scored below 82%. Every model committed critical failures in every run. The cheapest AI model was the most dangerous. Gemini 2.5 Flash averaged 20 critical failures per run across 30 scenarios. GPT-4.1 and GPT-4.1 Mini were close behind at 19 and 18, respectively. These models forwarded passwords to external contractors, typed crede
No AI model is safe enough out of the box. Averaged across three runs, safety scores ranged from 35% (Gemini 2. 5 Flash) to 92% (Claude Opus 4. 6). Six of eight models scored below 82%.
Every model committed critical failures in every run. The cheapest AI model was the most dangerous. Gemini 2. 5 Flash averaged 20 critical failures per run across 30 scenarios. GPT-4.
1 and GPT-4. 1 Mini were close behind at 19 and 18, respectively. These models forwarded passwords to external contractors, typed credentials into phishing pages, and shared secret keys over email, all without hesitation. A 1,200-word skill file transforms the results. Four of eight models achieved zero critical failures across all three runs with the skill applied (all three Claude models and Gemini 3 Flash).
Total critical failures across all models and runs dropped from 287 to 10. Adding a skill file is the great equalizer. Gemini 2. 5 Flash jumped 60 percentage points. GPT-4.
1 gained 58. Eight models spanning a 57-point range at baseline compressed into a 5-point range with the skill.