AgentReadyHomeAgent Listing

← seedance 3.0

seedance 3.0 — agentic threat model

5.6AIVSS 5.6 · Medium

Seedance 3.0 is a specialized AI video generator with low agentic autonomy, presenting primary risks around non-deterministic output generation, potential deepfake creation, and intellectual property concerns rather than systemic execution or infrastructure compromise.

OWASP AIVSS score rationale

AIVSS = (CVSS_Base + AARS) × Mitigation_Factor, where AARS = (10 − CVSS_Base) × (Factor_Sum / 10) × ThM
CVSS base 4.3AARS uplift 1.31Factor sum 2.3/10Threat ×1.0Mitigation ×1.0
Autonomy of Action
0.10
Goal-Driven Planning
0.10
Self-Modification
0.00
Dynamic Tool Use
0.10
Persistent Memory
0.20
Contextual Awareness
0.20
Dynamic Identity
0.00
Multi-Agent Interactions
0.00
Non-Determinism
0.80
Opacity & Reflexivity
0.80

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.

L1 · Foundation Models✓ mapped

Utilizes Bytedance's proprietary video generation foundation models (Seedance/Seedream). Primary threats include adversarial prompt injection to bypass safety filters, model extraction/stealing, and the generation of misaligned or harmful synthetic media (e.g., deepfakes).

L2 · Data Operations⚠ not certain from listing

Not certain from the listing — The model relies on massive video-text datasets for training. Key threats include copyright/IP infringement from training data, data lineage gaps, and potential leakage of proprietary training assets.

L3 · Agent Frameworks⚠ not certain from listing

Not certain from the listing — The system likely uses a straightforward prompt-to-generation pipeline rather than a complex agentic framework. Threats are limited to prompt manipulation bypassing input validation layers.

L4 · Deployment & Infrastructure⚠ not certain from listing

Not certain from the listing — Delivered via web and mobile apps, requiring heavy GPU infrastructure. Primary threats include API abuse, server-side resource exhaustion (denial of service via rendering requests), and insecure hosting environments.

L5 · Evaluation & Observability⚠ not certain from listing

Not certain from the listing — No details are provided on output monitoring. Threats include insufficient logging of generated content and a lack of robust automated guardrails to detect policy-violating or synthetic media outputs.

L6 · Security & Compliance (cross-cutting)⚠ not certain from listing

Not certain from the listing — As a closed-source commercial application, threats include non-compliance with emerging synthetic media regulations (such as watermarking mandates under the EU AI Act) and lack of transparent user data privacy controls.

L7 · Agent Ecosystem⚠ not certain from listing

Not certain from the listing — The tool operates as a standalone generator. Ecosystem threats are minimal, though integration into third-party video editing workflows could introduce minor supply-chain vulnerabilities.

MAESTRO — the 7-layer agentic threat-modeling framework (Cloud Security Alliance / Ken Huang).