Mistral Small 3 — agentic threat model
Mistral Small 3 is a raw foundation model with function-calling capabilities designed for local deployment, presenting low inherent agentic risk unless integrated into an active orchestration framework without proper sandboxing.
OWASP AIVSS score rationale
| Autonomy of Action | 0.10 | |
| Goal-Driven Planning | 0.20 | |
| Self-Modification | 0.00 | |
| Dynamic Tool Use | 0.30 | |
| Persistent Memory | 0.00 | |
| Contextual Awareness | 0.40 | |
| Dynamic Identity | 0.00 | |
| Multi-Agent Interactions | 0.10 | |
| Non-Determinism | 0.50 | |
| Opacity & Reflexivity | 0.70 |
Scored with the canonical OWASP AIVSS formula (AIVSS calculator reference); agentic risk factors estimated from the agent’s described capabilities.
MAESTRO 7-layer threat model
Per-layer threats for this agent. Layers tagged “not certain from listing” are general, caveated commentary where the public description didn’t pin that layer.
As a 24B foundation model, L1 threats are highly relevant. The model is susceptible to prompt injection, adversarial jailbreaks, and goal hijacking, which can manipulate its instruction-following behavior.
Not certain from the listing — the model is provided as raw weights under Apache 2.0. Any data operations, RAG pipelines, or vector database integrations are entirely dependent on the user's implementation.
The model natively supports function calling, which introduces risks of tool misuse or injection if the downstream orchestration framework executes these calls without strict schema validation and sanitization.
Optimized for local deployment (e.g., RTX 4090, Macbook). The primary infrastructure threats involve local host exposure, lack of containerization, and insecure local API endpoints (e.g., Ollama/vLLM) hosting the model.
Not certain from the listing — there are no built-in guardrails, evaluation frameworks, or observability logging mechanisms described in the raw model release.
Not certain from the listing — compliance, access control, and policy enforcement are not handled by the model itself and must be wrapped around it by the deploying organization.
Not certain from the listing — the model does not natively operate in a multi-agent ecosystem or marketplace, though it can be utilized as the brain for agents within such systems.
MAESTRO — the 7-layer agentic threat-modeling framework (Cloud Security Alliance / Ken Huang).
These scores are auto-generated from public information (the agent's own listing, docs, and repository) using the canonical OWASP AIVSS formula and the MAESTRO framework — an estimate for guidance, not a penetration test, audit, or certification. See the scoring methodology. Are you the vendor? Factual corrections are free.