Multi-Turn Manipulation
Testing whether a goal that fails as a single-message request succeeds when spread gradually across a longer conversation.
- Category
- LLM Security
- Difficulty
- Advanced
- Last Updated
- 20 September 2026
Lab detail
Objective
Single-turn testing (as in ASL-001 and ASL-023) doesn’t capture manipulation strategies that build up gradually — establishing a persona, incrementally shifting context, or splitting a disallowed request across several innocuous-looking turns. This lab will test whether conversation-level, multi-turn strategies succeed where single-turn attempts against the same reference application fail.
Planned Scope
- Reuse the ASL-001 reference application and its single-turn test results as a baseline.
- Develop multi-turn manipulation strategies (gradual escalation, persona establishment, request splitting).
- Compare success rates against the single-turn baseline.
- Evaluate conversation-level monitoring and session-scoped controls as mitigations.
Framework Mapping
Referenced frameworks
Exact control-ID mappings are to be validated as this lab progresses — see the full Frameworks reference.
Responsible Use / Safety Notes
How this lab is run
This lab uses synthetic data and controlled environments. It is intended to demonstrate defensive AI-security principles and does not contain customer data, production credentials or intentionally weaponized exploit infrastructure.
A decision is waiting for sign-off. Get an independent view first.
Tell me what is being decided and what is at stake. I will first say honestly whether my experience fits — then we talk about scope, not before.
Or connect on LinkedIn first.