Ai Angels — agentic threat model
AI Angels presents low agentic risk due to its lack of external tool execution or autonomous planning, but carries high privacy and reputational risk due to its storage of sensitive, intimate user conversations and personalized memory.
OWASP AIVSS score rationale
| Autonomy of Action | 0.20 | |
| Goal-Driven Planning | 0.10 | |
| Self-Modification | 0.30 | |
| Dynamic Tool Use | 0.10 | |
| Persistent Memory | 0.80 | |
| Contextual Awareness | 0.50 | |
| Dynamic Identity | 0.10 | |
| Multi-Agent Interactions | 0.00 | |
| Non-Determinism | 0.80 | |
| 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.
Not certain from the listing — likely uses fine-tuned text LLMs optimized for roleplay and emotional intelligence, alongside diffusion models for image/video generation. Vulnerable to jailbreaks bypassing safety filters or generating non-consensual/harmful imagery.
Not certain from the listing — requires a database or vector store to maintain long-term user conversation history ('memory'). Vulnerable to data leaks of highly sensitive, intimate user chats and profile preferences.
Not certain from the listing — uses a custom orchestration layer to manage dialogue state, personality profiles, and trigger image/video generation. Vulnerable to state manipulation or prompt injection that alters the companion's persona.
Not certain from the listing — likely hosted on cloud infrastructure with GPU acceleration for real-time inference and media generation. Vulnerable to standard web application exploits and unauthorized access to backend model APIs.
Not certain from the listing — monitoring is likely focused on user engagement and payment conversion rather than strict content alignment, given the NSFW nature. Vulnerable to undetected drift in bot behavior or abusive user inputs.
Not certain from the listing — claims a 'private, secure' experience but lacks details on encryption, data retention policies, or age verification mechanisms, which are critical for NSFW platforms.
The agent operates as a closed, standalone platform with no multi-agent marketplace or external ecosystem integrations described, minimizing external cascading risks.
MAESTRO — the 7-layer agentic threat-modeling framework (Cloud Security Alliance / Ken Huang).