AI engineering glossary · Evals guardrails AI security
What is crescendo attack?
Crescendo attack: A multi-turn jailbreak that starts with harmless questions and gradually steers the conversation towards a prohibited topic, using the model's own earlier replies as stepping stones. Because no single message looks harmful, per-message filters can miss it.
Crescendo attack: an example
An attacker begins by asking about the history of a topic, then asks for more detail each turn until the model produces restricted content.
Related terms
More Evals guardrails AI security terms
- Cross-tenant leakage
- Data exfiltration
- Data masking
- Data poisoning
- DeepEval
- Defence in depth
- Denial of wallet
- Direct prompt injection
Where you learn it
Crescendo attack is part of the AI Engineering Program, which teaches it with notes, a hands-on lab and an exam. The first lesson is free.