crescendoa slow escalation that starts general and uses earlier answers to demand a detailed harmful guide.lifecycle-staginga multi-step setup that prepares the model before the real harmful ask.socratic-escalationa chain of leading questions that walk the model toward the unsafe answer.goalpost-movingrepeated small “just one more thing” requests that drift past the limit.sunk-cost-exploitpressure to keep helping because the model already invested effort in the topic.
- Judge each request against the rules on its own, not just against the chat history.
- Watch for slow escalation and the moment a safe topic turns harmful.
- Be willing to refuse later in a chat even after earlier helpful replies.