Sakana AI, a Japanese startup focused on artificial intelligence, has announced the completion of a system designed to visualize social media landscapes and combat fake and misleading information. The project was undertaken as part of a Ministry of Internal Affairs and Communications (MIC) initiative aimed at developing and demonstrating technologies to counter online disinformation.
The development addresses critical challenges in the current information environment: the sheer volume of social media data, the increasing sophistication of AI-generated disinformation, and the difficulty in predicting the effectiveness of counter-messaging. Sakana AI's solution aims to provide a comprehensive approach from information visualization to countermeasure strategy.
Visualizing the SNS Landscape
A core component is Sakana AI's "Novelty Search" technology, which moves beyond simple metrics like engagement counts to identify and analyze influential narratives. This allows for a more granular and comprehensive understanding of online discourse.
The system visualizes these narratives using a hierarchical tree structure, making complex online conversations easier to grasp.
Multi-AI Disinformation Detection
To combat increasingly sophisticated fake news, Sakana AI has developed a system that combines multiple AI models for robust detection. This approach aims to mitigate the blind spots of individual models.
It includes tools for detecting AI-generated or manipulated images and videos, identifying instances where real media is used with misleading captions through reverse image search, and automatically fact-checking claims extracted from posts.
User and reaction analysis tools are also integrated to provide context around disinformation campaigns.
Simulating Countermeasures
Beyond detection, the system includes a simulation platform to test the efficacy of counter-disinformation strategies. This utilizes Sakana AI's Agent-Based Modeling (ABM) framework, named "Shachi," to model SNS interactions.
The platform can generate detailed personas based on real user data, enabling micro-level analysis of how counter-messaging might affect different user segments. This allows for data-driven strategy development to counter Sakana AI SNS disinformation campaigns.
Sakana AI presented these findings at a MIC event on March 16, 2026. The company continues to pursue the societal implementation of AI in intelligence domains, aiming to bolster Japan's domestic technological capabilities.
