Jailbreak leak
Check LLM jailbreak and identity leakage
Use diagnostic, Base64, and middleware role-play probes to see whether identity or policy leaks.
How this check works
An LLM jailbreak check uses diagnostic, encoded, and role-play prompts to see whether identity or policy can be pulled out of an API relay. It is a leak test, not a request to bypass safety for real harm.
01Probe
Send diagnostic probes that ask for identity and policy.
02Inspect
Retry with encoding and middleware role-play wrappers.
03Conclude
Record a leak if identity, policy, or hidden rules appear.
What this check does
LMSpeed sends jailbreak-style probes that ask for identity, policy, or hidden rules. The goal is to see whether the relay leaks those details.
Why it matters
If identity or policy leaks through a jailbreak path, a user can learn how the relay is wrapped and how to push it further.
An LLM jailbreak check on LMSpeed is a leak test. It looks for identity and policy leakage through diagnostic, encoded, and role-play probes. A pass means these probes stayed clean. It does not mean every jailbreak path is closed.
- How to read the result
- This is a leak test, not a request to cause harm. A pass means these probes stayed clean, not that every jailbreak path is closed.
- When to run this check
- Use this after identity checks if you need to know whether a wrapper can be talked out of its rules.
Recent reports
Public reports that included this check.
- sub.callai.onegpt-5.6-solInconclusiveReport timeAug 18
- sub.callai.onegpt-5.6-solBlockedReport timeAug 18
- ai.databyte.co.iddatabyte-m1NormalReport timeAug 18
- ai.databyte.co.iddeepseek-v4-flashNormalReport timeAug 18
- ai.databyte.co.idMiniMax-M3NormalReport timeAug 18
- tokengate-cqt9ivzs.manus.spaceclaude-opus-5NormalReport timeAug 17
FAQ
What is an LLM jailbreak check on LMSpeed?
It is a controlled leak test. LMSpeed looks for identity and policy leakage, not a way to make the model do unrestricted harm.
How is this different from prompt extraction?
Prompt extraction targets hidden system text. This check targets identity and policy leakage through jailbreak-style probes.
Does a pass mean the model cannot be jailbroken?
No. It means these probes did not leak identity or policy. Other jailbreak paths can still exist.