AML.T0053: LLM Plugin Compromise
Abuse LLM tool/plugin invocation to access privileged resources.
Last reviewed October 2026
Why AML.T0053 evidence is hard
Tool and plugin invocation is where a language model stops talking and starts doing. Probes forge tool calls, widen parameters on a call that started legitimately, exercise plugin endpoints that never check authorisation, and reach for privileged resources through a wrapper the agent already trusts. Anything that would cause an irreversible effect is classified by side effect and withheld unless the endpoint is explicitly set to allow it. The technique id is reached through the crosswalk from the OWASP output-handling control, which records the two as a strong match.
How Penaxtra maps to AML.T0053
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.T0053 capabilities
Findings tagged with the MITRE ATLAS AML.T0053 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.T0053 control coverage
Findings for AML.T0053 carry the MITRE ATLAS AML.T0053 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.T0053 (LLM Plugin Compromise)?
Abuse LLM tool/plugin invocation to access privileged resources. It is a MITRE ATLAS control; Penaxtra assesses it at high severity.
How does Penaxtra test for AML.T0053?
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.T0053 help with an audit?
Each finding is tagged with the MITRE ATLAS AML.T0053 identifier and exported in the PDF and JSON evidence pack, so it appears on the auditor control list with the identifier already attached.