SafeGuard: A Multi-Agent Perception-Reasoning Framework for Social-Risk AI-Generated Video Detection

2026-07-03

AI Summary

This work addresses the challenge that existing AI-generated video detection methods struggle to identify subtle anomalies in highly realistic deepfakes—such as violations of physical laws, structural inconsistencies, or social logic—primarily due to the disconnect between perceptual analysis and semantic reasoning. To bridge this gap, we propose SafeGuard, a novel multi-agent framework that introduces collaborative multi-agent mechanisms into deepfake detection for the first time. SafeGuard integrates hierarchical perception solvers to extract fine-grained forensic cues with self-reflective semantic verifiers that ensure reasoning adheres to physical and social plausibility, thereby constructing interpretable evidence chains. We also introduce SafeVid, the first benchmark dataset focused on socially risky deepfake scenarios. Experiments demonstrate that SafeGuard achieves an 18.7% accuracy improvement on SafeVid and significantly outperforms state-of-the-art methods across four public datasets, substantially enhancing generalization in detecting high-fidelity, socially hazardous synthetic videos.

Abstract

As video generation paradigms evolve from localized manipulation to full-scene synthesis, AI-generated video detection becomes increasingly challenging, as forgeries exhibit coherent global structure and high perceptual realism. However, existing benchmarks are biased toward perceptual fidelity and primarily evaluate detectors based on perceptual artifacts, providing limited coverage of scenarios that require reasoning about violations of physical laws, structural coherence, or social logic. This dataset bias shapes current approaches and results in a Perception-Reasoning Gap: artifact-centric models capture low-level statistical irregularities yet lack semantic inference, whereas vision-language models perform semantic reasoning but remain insensitive to fine-grained forensic cues. To bridge this gap, we propose SafeGuard, a multi-agent framework that enables collaborative specialization between forensic perception and semantic reasoning. A hierarchical perceptual solver extracts fine-grained forensic evidence, while a self-reflective verifier enforces consistency between semantic inference and physical plausibility, forming an interpretable evidence chain. To support evaluation, we introduce SafeVid, a novel AI-generated video detection benchmark comprising 20K videos spanning 10 social risk categories, designed to evaluate physical plausibility, structural consistency, and the rationality of social behaviors. Extensive experiments demonstrate the generalization of SafeGuard, improving accuracy on SafeVid by +18.7% and consistently outperforming prior methods across four public benchmarks.

Problem

• AI-generated video detection
• perception-reasoning gap
• social-risk scenarios
• physical plausibility
• semantic inconsistency

Innovation

• multi-agent framework
• perception-reasoning gap
• forensic perception
• semantic reasoning
• AI-generated video detection