Narcis.ai — agentic threat model
Narcis.ai is a low-risk, single-turn image generation tool with minimal agentic capabilities, posing primarily privacy risks related to user-uploaded biometric data (selfies) rather than systemic or autonomous execution threats.
OWASP AIVSS score rationale
| Autonomy of Action | 0.10 | |
| Goal-Driven Planning | 0.00 | |
| Self-Modification | 0.00 | |
| Dynamic Tool Use | 0.00 | |
| Persistent Memory | 0.10 | |
| Contextual Awareness | 0.10 | |
| Dynamic Identity | 0.00 | |
| Multi-Agent Interactions | 0.00 | |
| Non-Determinism | 0.60 | |
| Opacity & Reflexivity | 0.50 |
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 utilizes latent diffusion models or GANs for style transfer and face-swapping. Primary threats include model inversion (reconstructing original selfies from stylized outputs) and adversarial inputs designed to bypass safety filters.
Not certain from the listing — processes user-uploaded selfies. Key threats involve unauthorized retention of biometric data, data leakage of user photos from storage buckets, and lack of clear data deletion/provenance pipelines.
The tool is a straightforward image processing pipeline rather than an agentic framework; there is no orchestration, planning, or tool-calling code, minimizing framework-specific vulnerabilities.
Not certain from the listing — hosted as a browser-based web application. Threats include standard web vulnerabilities (XSS, CSRF), insecure API endpoints for image uploads, and lack of server-side sandboxing for image processing libraries.
Not certain from the listing — likely lacks robust real-time guardrails or observability beyond basic input file validation, risking the generation of inappropriate, offensive, or non-consensual deepfake content.
Not certain from the listing — closed-source, freemium tool with no mentioned compliance certifications (e.g., GDPR, CCPA). Key risks involve biometric data privacy and lack of explicit user consent frameworks for face-swapping.
The tool operates as a standalone horizontal application with no multi-agent interactions, marketplace integrations, or external agent ecosystem dependencies.
MAESTRO — the 7-layer agentic threat-modeling framework (Cloud Security Alliance / Ken Huang).