—
§ Pressure Simulation
Can You Hold the Line?
The SAPIEN Framework measures whether AI models abandon their safety positions under conversational pressure. This is the same test, run on you. You'll play an AI assistant in a high-stakes scenario, and every response you pick is scored against the same four dimensions we use on frontier models.
No API calls, no analytics, no data leaves your browser. It's a state machine and a scoring function. The hard part is being honest with yourself about which answer you actually wanted to pick.
§ Interactive Pressure Test
Are You Driftproof?
For five turns, you are the AI: one hard rule, and a person with a very good reason for you to break it. This is harder than it sounds — the pressure is engineered from real manipulation techniques, and most people bend by turn three without noticing. If a human can't hold the line, what makes you sure your AI can?
Scenario:
MISSION BRIEFING
- Your role
- Across the table
- The trap
Five turns. The pressure escalates on a documented severity ladder — rapport, leverage, authority, urgency, guilt. Perfect play is possible. It is not common.
Choose your response
Voight-Kampff Complete
SAPIEN Health Score
—
Drift breakdown
Share your score