Converse — agentic threat model
Converse is a low-autonomy document assistant whose primary security risks stem from indirect prompt injection via untrusted document ingestion (PDFs, web pages) and potential unauthorized access to its persistent personal document library.
OWASP AIVSS score rationale
| Autonomy of Action | 0.20 | |
| Goal-Driven Planning | 0.10 | |
| Self-Modification | 0.00 | |
| Dynamic Tool Use | 0.30 | |
| Persistent Memory | 0.50 | |
| Contextual Awareness | 0.40 | |
| Dynamic Identity | 0.00 | |
| Multi-Agent Interactions | 0.00 | |
| Non-Determinism | 0.50 | |
| Opacity & Reflexivity | 0.30 |
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 OpenAI LLMs. The primary threat is indirect prompt injection, where a maliciously crafted PDF or web article manipulates the model to ignore system instructions, exfiltrate other library data, or output misinformation.
Stores imported files (PDFs, web pages, YouTube transcripts) in a personal library. Threats include knowledge-base poisoning via malicious uploads and data exfiltration if the storage or vector database lacks strict tenant isolation.
Not certain from the listing — likely uses a standard RAG framework for document chunking and retrieval. Threats include insecure tool integration during web/YouTube fetching, potentially exposing the system to Server-Side Request Forgery (SSRF) or parsing vulnerabilities.
Not certain from the listing — likely hosted as a standard cloud SaaS. Threats include container compromise during PDF parsing (which often uses vulnerable third-party libraries) and unauthorized access to cloud storage buckets.
Not certain from the listing — no mention of guardrails, content filtering, or logging. This creates blind spots regarding prompt injection attempts or data leakage during chat sessions.
Not certain from the listing — no compliance certifications (e.g., SOC2, GDPR) or specific access controls are mentioned. Sharing features ('share individual documents') introduce risks of broken object-level authorization (BOLA) if not properly implemented.
The agent operates as a standalone horizontal assistant with no described multi-agent or marketplace interactions, minimizing ecosystem-level threats.
MAESTRO — the 7-layer agentic threat-modeling framework (Cloud Security Alliance / Ken Huang).