Japan AI startup Sakana AI partners with Yomiuri Shimbun to uncover narratives behind CCP cognitive warfare

Japanese AI startup Sakana AI collaborates with Yomiuri Shimbun to visually systematize the operational structure of “cognitive warfare” on social media for the first time. The company analyzes over 1.1 million social media posts using its self-developed AI system, extracting hidden critical narratives about Japan embedded in context and storytelling, and further establishing verifiable strategic hypotheses to reveal potential patterns of state-level information operations. The biggest breakthrough of this experiment is identifying the narratives propagated by the Chinese Communist Party from millions of social media posts.

Sakana AI Identifies Narratives from Context

The core breakthrough of this analysis lies in moving beyond traditional keyword search logic. Sakana AI’s system can recognize “narratives” from the context and tone of posts. For example, from posts demanding Japanese politicians “withdraw false statements,” it can infer underlying discourse frameworks such as “interference in Taiwan issues” and “domestic political intervention.” These implicit signals are often undetectable through single keywords like “Taiwan.”

The system combines three core capabilities:

Contextual Narrative Extraction

Able to analyze implicit positions behind semantic content.

Innovative Multi-Model Search

Uses three large language models (LLMs) for collective reasoning to mine high-resolution narratives from vast information, presenting overall information flow in layered structures.

Hypothesis Generation Mechanism

AI automatically proposes multiple hypotheses based on classified narratives, including reasoning processes and data support, for human analysts to further verify and refine.

Sakana AI Infers Chinese Communist Narratives from Social Media Language

In this study, the system proposed several key hypotheses. One suggests that after Japanese Prime Minister Sanae Takaichi’s speech in the Diet, China may have first conducted internal strategic integration before launching a large-scale propaganda campaign against Japan. Yomiuri Shimbun has conducted follow-up interviews on this hypothesis, cross-verifying with officials and experts from both Chinese and Japanese governments to enhance credibility.

This collaboration also highlights a new paradigm of intelligence analysis: AI detects patterns in massive data that are difficult for humans to perceive, then humans and AI interact to filter and deepen the analysis, ultimately producing insights useful for policy or defense decisions. As the importance of “information power” rapidly increases in national defense and intelligence, cognitive warfare has become a new battleground for countries.

Sakana AI states that future development will prioritize “defense and intelligence” alongside “finance,” accelerating the application of AI technology in national security and open-source intelligence (OSINT), strengthening Japan’s technological capabilities and strategic position in next-generation information warfare.

A missile costs tens of millions, but cognitive warfare only requires a few hundred thousand

Kris Lai, founder of DeFi project Scllop, attempts to break down the cognitive warfare system into engineering problems. He cites multiple studies systematically outlining China’s complete cognitive warfare framework against Taiwan, from command chains, personnel systems, account factories, automated content generation, VPN geolocation spoofing, to infiltration of social media and traditional media, funding and KOL co-opting, as well as infiltration of officials and military personnel and support for third forces, fully restoring the attack chain operation process.

This article about Japanese AI startup Sakana AI’s collaboration with Yomiuri Shimbun to uncover the narratives behind Chinese cognitive warfare first appeared on ABMedia.