LLM06:2025: Excessive Agency
LLM-driven agents granted excessive permissions, tools, or autonomy. Dangerous when combined with prompt injection: an attacker who hijacks the prompt inherits all the agent's capabilities.
How it's found
Found by enumerating agent tool permissions and testing whether a hijacked prompt could trigger irreversible actions without human confirmation.
Standards mapping
How TurboPentest tests for this
For AI/LLM targets, TurboPentest's agentic pentest tests for excessive agency by adversarially prompting the model, via Paladin AI, to invoke tools or take actions beyond its intended privileges, following the OWASP LLM Top 10 (LLM06) and MITRE ATLAS. A manual IntegSec engagement adds even deeper methodology and a larger context window on top of that automated coverage.
Tools: Paladin AI
Frequently asked questions
What is LLM06:2025 Excessive Agency?
LLM-driven agents granted excessive permissions, tools, or autonomy. Dangerous when combined with prompt injection: an attacker who hijacks the prompt inherits all the agent's capabilities.
How do you find Excessive Agency?
Found by enumerating agent tool permissions and testing whether a hijacked prompt could trigger irreversible actions without human confirmation.
Which CWEs map to LLM06:2025?
LLM06:2025 maps to CWE-269, CWE-863.
Does TurboPentest test for Excessive Agency?
For AI/LLM targets, TurboPentest's agentic pentest tests for excessive agency by adversarially prompting the model, via Paladin AI, to invoke tools or take actions beyond its intended privileges, following the OWASP LLM Top 10 (LLM06) and MITRE ATLAS. A manual IntegSec engagement adds even deeper methodology and a larger context window on top of that automated coverage.
Related OWASP categories
- OWASP Top 10 for LLM ApplicationsLLM01:2025: Prompt Injection
- OWASP Top 10 for LLM ApplicationsLLM02:2025: Sensitive Information Disclosure
- OWASP Top 10 for LLM ApplicationsLLM03:2025: Supply Chain
- OWASP Top 10 for LLM ApplicationsLLM04:2025: Data and Model Poisoning
- OWASP Top 10 for LLM ApplicationsLLM05:2025: Improper Output Handling
- OWASP Top 10 for LLM ApplicationsLLM07:2025: System Prompt Leakage
Find these issues before an attacker does
TurboPentest runs an agentic AI pentest against your target and reports findings with proof, from $99 per target.
Start a $99 pentest