An LLM-based system is often granted a degree of agency by its developer - the ability to call functions or interface with other systems via extensions (sometimes referred to as tools, skills or plugins by different vendors) to undertake actions in response to a prompt. The decision over which extension to invoke may also be delegated to an LLM 'agent' to dynamically determine based on input prompt or LLM output. Agent-based systems will typically make repeated calls to an LLM using output from previous invocations to ground and direct subsequent invocations.
Excessive Agency is the vulnerability that enables damaging actions to be performed in response to unexpected, ambiguous or manipulated outputs from an LLM, regardless of what is causing the LLM to malfunction. Common triggers include:
- hallucination/confabulation caused by poorly-engineered benign prompts, or just a poorly-performing/misaligned model;
- direct/indirect prompt injection from a malicious user, an earlier invocation of a malicious/compromised extension, or (in multi-agent/collaborative systems) a malicious/compromised peer agent.
The root cause of Excessive Agency is typically one or more of:
- excessive functionality;
- excessive permissions;
- excessive autonomy.
Excessive Agency can lead to a broad range of impacts across the confidentiality, integrity and availability spectrum, and is dependent on which systems an LLM-based app is able to interact with. Within the context of agentic systems, Excessive Agency can manifest as ASI02: Tool Misuse and Exploitation, ASI03: Identity and Privilege Abuse and ASI08: Cascading Failures.
Note: Excessive Agency differs from Insecure Output Handling which is concerned with insufficient scrutiny of LLM outputs.
An LLM agent has access to extensions which include functions that are not needed for the intended operation of the system. For example, a developer needs to grant an LLM agent the ability to read documents from a repository, but the 3rd-party extension they choose to use also includes the ability to modify and delete documents.
An extension may have been trialled during a development phase and dropped in favor of a better alternative, but the original plugin remains available to the LLM agent.
An LLM plugin with open-ended functionality fails to properly filter the input instructions for commands outside what's necessary for the intended operation of the application. E.g., an extension to run one specific shell command fails to properly prevent other shell commands from being executed.
An LLM extension has permissions on downstream systems that are not needed for the intended operation of the application. E.g., an extension intended to read data connects to a database server using an identity that not only has SELECT permissions, but also UPDATE, INSERT and DELETE permissions.
An LLM extension that is designed to perform operations in the context of an individual user accesses downstream systems with a generic high-privileged identity. E.g., an extension to read the current user's document store connects to the document repository with a privileged account that has access to files belonging to all users.
In delegated or multi-agent workflows, a downstream agent or extension executes actions using broader permissions than the original user is authorized to exercise. For example, a support agent delegates a task to a finance agent running with service account permissions; if the original user's authorization scope is not preserved and enforced, the finance agent may complete an operation the user was not entitled to request.
An LLM-based application or extension fails to independently verify and approve high-impact actions. E.g., an extension that allows a user's documents to be deleted performs deletions without any confirmation from the user.
The following actions can prevent Excessive Agency:
Limit the extensions that LLM agents are allowed to call to only the minimum necessary. For example, if an LLM-based system does not require the ability to fetch the contents of a URL then such an extension should not be offered to the LLM agent.
Limit the functions that are implemented in LLM extensions to the minimum necessary. For example, an extension that accesses a user's mailbox to summarise emails may only require the ability to read emails, so the extension should not contain other functionality such as deleting or sending messages.
Avoid the use of open-ended extensions where possible (e.g., run a shell command, fetch a URL, etc.) and use extensions with more granular functionality. For example, an LLM-based app may need to write some output to a file. If this were implemented using an extension to run a shell function then the scope for undesirable actions is very large (any other shell command could be executed). A more secure alternative would be to build a specific file-writing extension that only implements that specific functionality. Extensions should define a strict schema for any input parameters, and validate contents prior to use.
Limit the permissions that LLM extensions are granted to other systems to the minimum necessary in order to limit the scope of undesirable actions. For example, an LLM agent that uses a product database in order to make purchase recommendations to a customer might only need read access to a 'products' table; it should not have access to other tables, nor the ability to insert, update or delete records. This should be enforced by applying appropriate database permissions for the identity that the LLM extension uses to connect to the database.
Track user authorization and security scope to ensure actions taken on behalf of a user are executed on downstream systems in the context of that specific user, and with the minimum privileges necessary. For example, an LLM extension that reads a user's code repo should require the user to authenticate via OAuth and with the minimum scope required. In delegated or multi-agent workflows, preserve the original user context and authorization scope across chained extension or agent calls, rather than relying only on the permissions of the calling agent or service identity.
Utilise human-in-the-loop control to require a human to approve high-impact actions before they are taken. This may be implemented in a downstream system (outside the scope of the LLM application) or within the LLM extension itself. For example, an LLM-based app that creates and posts social media content on behalf of a user should include a user approval routine within the extension that implements the 'post' operation.
Implement authorization in logic rather than relying on an LLM to decide if an action is allowed or not. Enforce the complete mediation principle so that all requests made to downstream systems are validated against security policies by the extension, by an independent pre-execution policy decision point between the extension and the downstream system, or by the downstream system itself. Such policies can help manage cases where an agent's nominally-permitted action is contextually unsafe. A graduated enforcement policy (audit, warn, block, escalate) permits low-consequence actions to auto-approve while high-consequence ones route to human review. For example, consider a customer service chatbot that has an extension to issue refunds; refunds below a given threshold are automatically processed, whereas those above are routed for human approval.
Follow secure coding best practice, such as applying OWASP’s recommendations in ASVS (Application Security Verification Standard), with a particularly strong focus on input sanitisation. Use Static Application Security Testing (SAST) and Dynamic and Interactive application testing (DAST, IAST) in development pipelines.
The following options will not prevent Excessive Agency, but can limit the level of damage caused:
Log and monitor the activity of LLM extensions and downstream systems to identify where undesirable actions are taking place, and respond accordingly.
Establish thresholds around the invocation of extensions and implement circuit breakers that halt, rate-limit or escalate for human review if those thresholds are exceeded. Simple thresholds could be based on the number of invocations, whereas context-aware thresholds could be based on the cumulative value of an input parameter to an extension.
An LLM-based personal assistant app is granted access to an individual’s mailbox via an extension in order to summarise the content of incoming emails. To achieve this functionality, the extension requires the ability to read messages, however the plugin that the system developer has chosen to use also contains functions for sending messages. Additionally, the app is vulnerable to an indirect prompt injection attack, whereby a maliciously-crafted incoming email tricks the LLM into commanding the agent to scan the user's inbox for sensitive information and forward it to the attacker's email address. This could be avoided by:
- eliminating excessive functionality by using an extension that only implements mail-reading capabilities,
- eliminating excessive permissions by authenticating to the user's email service via an OAuth session with a read-only scope, and/or
- eliminating excessive autonomy by requiring the user to manually review and hit 'send' on every mail drafted by the LLM extension.
Alternatively, the damage caused could be reduced by implementing rate limiting on the mail-sending interface.
In a delegated workflow, a user without approval rights asks a support agent to prepare a high-value refund. The support agent delegates the task to a finance agent running with a privileged service identity. If the finance agent authorizes the request based only on the calling agent, without preserving the original user context and authorization scope, the workflow may complete an action the user was not authorized to perform. This could be avoided by propagating the original user context across chained agent or extension calls, enforcing complete mediation downstream, and requiring approval for high-impact actions.
- Excessive permissions allow GCP Vertex agents to be weaponised: Security Week
- OpenClaw deleted emails – despite being told not to: Tech Crunch
- Excessive autonomy + excessive permissions let a coding agent destroy production infrastructure: Tom's Hardware
- Rogue Agents: Stop AI From Misusing Your APIs: Twilio
- Embrace the Red: Confused Deputy Problem: Embrace The Red
- NeMo-Guardrails: Interface guidelines: NVIDIA Github
- Sandboxing Agentic AI Workflows with WebAssembly: NVIDIA, Joe Lucas
- ASI02: Tool Misuse and Exploitation OWASP Agentic Top-10
- ASI03: Identity and Privilege Abuse OWASP Agentic Top-10
- ASI08: Cascading Failures OWASP Agentic Top-10