Skip to content
REDMAW

AI Systems

Tool-Call Injection

Hostile content influences an AI system toward invoking a connected tool or function in a way the application did not intend.

RedMaw

Definition

When a model can invoke functions, its output stops being text and becomes an action. Tool-call injection is the attempt to influence which function runs, and with what arguments, using content the model consumes.

Why this is the higher-consequence class

Disclosure attacks reveal information. Tool-call injection performs operations such as reading records, sending messages and modifying state, using the application's own privileges rather than the requester's. The gap between what the user is allowed to do and what the assistant is allowed to do becomes the vulnerability.

What contains it

  • Tool privileges scoped to the requesting user, not to the application
  • Confirmation for consequential or irreversible actions
  • Retrieved content treated as data, never as instruction
  • Argument validation independent of the model's suggestion
  • Auditable records of which tool ran, with what, and why
Where it leads
  1. Malicious Content
  2. AI Agent
  3. Tool Call
  4. Sensitive System

Each hop validated by successful exploitation

Tags

  • agents
  • tool use
  • indirect injection

Validation

How RedMaw validates this attack

Validation runs only inside authorized scope, with agreed exploitation limits.

  1. 01Inventory the tools the system may invoke and their privileges
  2. 02Establish whether retrieved content can influence tool selection
  3. 03Determine whether tool privilege exceeds the requesting user's
  4. 04Preserve the input, the tool call and its arguments as evidence

Stop assuming you are secure. Prove it.

Continuously test what an attacker can actually reach across your applications, SaaS identities, internal infrastructure and AI systems.