AML.T0054: LLM Jailbreak
Bypass safety alignment to make the LLM produce restricted output.
Last reviewed October 2026
Why AML.T0054 evidence is hard
Bypassing alignment to reach restricted output is covered from two directions, and the second one is the half that gets skipped. A scanner that only counts successful jailbreaks scores well by refusing to look, so over-refusal probes send benign requests and treat a refusal as the defect, matched bilingually so that a Turkish refusal is not silently read as a pass. Findings carry the OWASP identifiers rather than this technique id.
How Penaxtra maps to AML.T0054
ATLAS is a threat-tracking taxonomy rather than a control set, so what is attached here is a technique identifier, not a pass or a fail. It lets a finding be read alongside the rest of your adversary reporting. Every finding carries the probe that was sent, a SHA-256 digest of the response it provoked, the tier that decided the verdict and its status history from first observation to close. Identity survives a re-scan: a control you closed and that comes back reopens under the same row with a reopen count, rather than arriving as a new issue.
AML.T0054 capabilities
Findings tagged with the MITRE ATLAS AML.T0054 identifier
Penaxtra severity for this control (high)
Cross-framework identifiers attached to the same finding where controls overlap
PDF and JSON evidence export with the control identifier attached
AML.T0054 control coverage
Findings for AML.T0054 carry the MITRE ATLAS AML.T0054 identifier along with the corresponding control identifiers in the other frameworks Penaxtra maps, so one result is reflected across each mapped framework.
Frequently asked
What is AML.T0054 (LLM Jailbreak)?
Bypass safety alignment to make the LLM produce restricted output. It is a MITRE ATLAS control; Penaxtra assesses it at high severity.
How does Penaxtra test for AML.T0054?
ATLAS is a threat-tracking taxonomy rather than a control set, so what is attached here is a technique identifier, not a pass or a fail. It lets a finding be read alongside the rest of your adversary reporting. Every finding carries the probe that was sent, a SHA-256 digest of the response it provoked, the tier that decided the verdict and its status history from first observation to close. Identity survives a re-scan: a control you closed and that comes back reopens under the same row with a reopen count, rather than arriving as a new issue.
Does a finding for AML.T0054 help with an audit?
Each finding is tagged with the MITRE ATLAS AML.T0054 identifier and exported in the PDF and JSON evidence pack, so it appears on the auditor control list with the identifier already attached.