Seedance 2.0 app — agentic threat model

Uses advanced multimodal AI models to interpret text and images for video/audio generation. Primary threats include adversarial prompt injection to bypass safety filters (generating NSFW, copyrighted, or harmful content) and model extraction if proprietary weights are exposed.

Not certain from the listing — No details are provided regarding training data, RAG, or vector stores. General threats include training data poisoning, copyright/IP infringement from training datasets, and lack of data lineage for generated assets.

Not certain from the listing — The application appears to function as a direct generation pipeline rather than a complex agentic framework. General threats include insecure orchestration of the video/audio rendering pipeline and potential command injection via malformed input metadata.

Not certain from the listing — No hosting or infrastructure details are provided, though 2K video generation implies heavy GPU utilization. General threats include GPU resource exhaustion (denial of service) and container/host compromise on rendering servers.

Not certain from the listing — No mention of content moderation guardrails, logging, or output evaluation. General threats include the lack of automated detection for deepfakes, CSAM, or malicious synthetic media generated by the platform.

Not certain from the listing — No compliance certifications (e.g., SOC2, GDPR) or identity management details are cited. General threats include lack of robust user access controls and audit logs to trace the origin of malicious synthetic media.

Not certain from the listing — No multi-agent or marketplace interactions are described. General threats include the potential for this tool to be integrated into external automated botnets or misinformation pipelines to generate high-volume synthetic media.