Kimi AI — agentic threat model
Kimi AI presents a moderate agentic risk profile, primarily driven by its high-capacity data ingestion (200k characters) and real-time web search capabilities, which expose it to indirect prompt injection and data exfiltration threats, while its lack of autonomous execution limits direct physical or system-level impacts.
OWASP AIVSS score rationale
| Autonomy of Action | 0.30 | |
| Goal-Driven Planning | 0.20 | |
| Self-Modification | 0.10 | |
| Dynamic Tool Use | 0.40 | |
| Persistent Memory | 0.30 | |
| Contextual Awareness | 0.70 | |
| Dynamic Identity | 0.10 | |
| Multi-Agent Interactions | 0.10 | |
| Non-Determinism | 0.60 | |
| 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.
Uses Moonshot AI's proprietary LLM. Vulnerable to prompt injection, adversarial inputs (especially via long text/images), and model reprogramming.
Processes up to 200k characters of user-uploaded data (text, code, images, audio) and integrates real-time web data. Vulnerable to data exfiltration via prompt injection, and indirect prompt injection via web search or uploaded files.
Not certain from the listing — the orchestration framework for tool execution (web search, audio transcription, file parsing) is proprietary; vulnerabilities could lead to insecure tool calling or SSRF via web search.
Not certain from the listing — hosted as a closed-source cloud service by Moonshot AI; infrastructure security, sandboxing of file parsers, and network isolation are managed internally without public details.
Not certain from the listing — monitoring, guardrails, and abuse detection mechanisms are proprietary and not detailed in the public directory.
Not certain from the listing — compliance certifications (e.g., GDPR, local Chinese data regulations) and enterprise access controls are not specified in the basic directory listing.
Not certain from the listing — Kimi AI operates primarily as a standalone assistant; there is no explicit mention of multi-agent marketplace interactions or third-party agent integrations.
MAESTRO — the 7-layer agentic threat-modeling framework (Cloud Security Alliance / Ken Huang).