Skip to content
← Back to Glossary
AI Security

Jailbreak

Bypassing an LLM's content safety filters with creative prompting.

Jailbreak attacks craft special prompts to bypass LLM safety filters — using roleplay scenarios, encoding tricks, or hypothetical framing. While similar to prompt injection, jailbreaks specifically target safety guardrails. Defense-in-depth with input/output monitoring is essential.

Related Terms

See if your app is vulnerable

Vezraa checks for related security issues. Scan in 25 seconds.

Jailbreak — Security Glossary | Vezraa