seedance 2 — agentic threat model
Seedance 2.0 is primarily a generative video model with low agentic autonomy, posing minimal systemic risk but presenting notable risks related to deepfake generation, prompt injection, and intellectual property concerns.
OWASP AIVSS score rationale
| Autonomy of Action | 0.10 | |
| Goal-Driven Planning | 0.20 | |
| Self-Modification | 0.00 | |
| Dynamic Tool Use | 0.00 | |
| Persistent Memory | 0.10 | |
| Contextual Awareness | 0.20 | |
| Dynamic Identity | 0.00 | |
| Multi-Agent Interactions | 0.00 | |
| Non-Determinism | 0.70 | |
| 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 Seedance 2.0 foundation model for video synthesis. Primary threats include adversarial prompt injection to bypass safety filters (generating NSFW or deepfake content), model stealing, and output misalignment.
Not certain from the listing — Data operations likely involve massive video-image pre-training datasets. Key threats include data poisoning of the training pipeline, copyright/provenance gaps, and potential leakage of proprietary training data.
Not certain from the listing — Orchestration is likely limited to sequential scene generation (multi-shot storytelling). Threats include prompt injection manipulating the story planner to generate unintended narrative sequences.
Not certain from the listing — Hosted on cloud GPU infrastructure. Threats include API abuse, resource exhaustion (DoS) due to heavy video rendering demands, and unauthorized access to user-generated video assets.
Not certain from the listing — No details on output guardrails or observability. Threats include a lack of automated deepfake detection, watermarking bypasses, and insufficient logging of malicious prompt attempts.
Not certain from the listing — Closed-source, paid model. Threats include non-compliance with synthetic media regulations (such as the EU AI Act's watermarking requirements) and lack of transparent data privacy policies.
Not certain from the listing — No multi-agent ecosystem or marketplace interactions are described. Threats are limited to unauthorized third-party API wrappers or integrations.
MAESTRO — the 7-layer agentic threat-modeling framework (Cloud Security Alliance / Ken Huang).