seedance 3.0 — agentic threat model
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
| 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.
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).
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.
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.
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.
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.
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.
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).