China deploys AI for large-scale public opinion manipulation against Japan
China has fully deployed AI-generated content for public opinion manipulation against Japan, structurally testing the defense capabilities of Japanese SNS platforms and democracy. The shift from traditional human-intensive influence operations to AI automation fundamentally changes the cost structure of information warfare, making the risk of overwhelming defensive capabilities a reality.
── Understand in 3 points ─────────
- • China's AI influence operations leverage Large Language Models (LLMs) and have acquired the ability to generate large volumes of natural-sounding posts in Japanese. Since 2024, OpenAI has reported disrupting multiple China-linked influence operation networks.
- • According to Meta's 2024 Threat Report, the China-originated "Spamouflage" network is the world's largest influence operation network, active on over 50 platforms. Japanese content is also among its targets.
- • Japan lacks comprehensive platform regulation laws equivalent to the EU's Digital Services Act (DSA) or Australia's Online Safety Act. Disinformation countermeasures in 2024 remain primarily within the framework of the Ministry of Internal Affairs and Communications' "voluntary initiatives."
── NOW PATTERN ─────────
China's AI public opinion weapons are offensive tools designed to seize "narrative hegemony," utilizing the "platform dominance" structure of SNS platforms as an attack vector, and Japan's "institutional decay" (or more accurately, lack of institutional development) in disinformation countermeasures provides a defensive vulnerability, thus functioning in a three-layered structure.
── Probability and Response ──────
🟡 Base 55% — Ministry of Internal Affairs and Communications' bill submission schedule, scale of Ministry of Defense's cognitive warfare budget, concrete progress in Japan-Taiwan information sharing, changes in content of major platforms' transparency reports
🟢 Optimistic 15% — Occurrence of a large-scale AI influence operation exposure incident, bipartisan bill submission, movement towards organizational integration of the Digital Agency and Ministry of Defense, official announcement of Japan-Taiwan cognitive warfare cooperation
🔴 Pessimistic 30% — Delay in bill submission, announcement of moderation cuts by platform operators, rapid escalation of military tensions in the Taiwan Strait, occurrence of large-scale distribution of Japanese deepfakes
📡 The Signal — What Happened
Why it matters: China has fully deployed AI-generated content for public opinion manipulation against Japan, structurally testing the defense capabilities of Japanese SNS platforms and democracy. The shift from traditional human-intensive influence operations to AI automation fundamentally changes the cost structure of information warfare, making the risk of overwhelming defensive capabilities a reality.
- Technology — China's AI influence operations leverage Large Language Models (LLMs) and have acquired the ability to generate large volumes of natural-sounding posts in Japanese. Since 2024, OpenAI has reported disrupting multiple China-linked influence operation networks.
- Scale — According to Meta's 2024 Threat Report, the China-originated "Spamouflage" network is the world's largest influence operation network, active on over 50 platforms. Japanese content is also among its targets.
- Institutions — Japan lacks comprehensive platform regulation laws equivalent to the EU's Digital Services Act (DSA) or Australia's Online Safety Act. Disinformation countermeasures in 2024 remain primarily within the framework of the Ministry of Internal Affairs and Communications' "voluntary initiatives."
- Geopolitics — As tensions rise over the Taiwan Strait, China's information warfare against Japan has the strategic objectives of "weakening the Japan-U.S. alliance" and "neutralizing Japan in a Taiwan contingency." The Ministry of Defense's 2025 Defense White Paper for the first time dedicated a separate chapter to threats in the cognitive domain.
- Technology — "Multimodal influence operations" combining generative AI deepfake videos, voice cloning, and AI translation have rapidly advanced since 2024. TikTok and X (formerly Twitter) are particularly becoming major distribution channels.
- Economics — The cost of influence operations using AI-generated content is estimated to be about 1/100th of traditional methods using human operators. It has the capacity to automatically generate and distribute tens of thousands of posts per day.
- Defense — NATO's Strategic Communications Centre of Excellence (StratCom) warned in its 2024 report that AI-powered influence operations are "rapidly improving their ability to evade detection."
- Policy — The Japanese government published its "Summary of Countermeasures against False and Misleading Information" in December 2024, but it has not led to legally binding regulations. The Active Cyber Defense Bill is being deliberated in the 2025 ordinary Diet session, but its scope for information warfare countermeasures is limited.
- Alliances — The Five Eyes alliance countries are strengthening information sharing regarding China's AI influence operations, but Japan is not a formal member, leading to structural limitations in the depth of information sharing.
- Platforms — After Elon Musk's acquisition of X (formerly Twitter), content moderation capabilities significantly declined, with analyses indicating that the detection and deletion rate of bot accounts decreased by approximately 60% compared to before 2023.
- Public Opinion — In NHK's 2025 public opinion survey, 68% of Japanese respondents answered that "false information on the internet is increasing," but only 23% said they "can distinguish false information themselves."
- Military — The Strategic Support Force (SSF) of the Chinese People's Liberation Army (PLA), reorganized into the Information Support Force in 2024, integrates cyber warfare, electronic warfare, and psychological warfare, incorporating AI public opinion manipulation as part of its military doctrine.
The history of China's information warfare against Japan dates back to the pre-internet era. However, with the practical deployment of AI, this struggle has entered a qualitatively different phase. To understand its structural background, three historical contexts must be overlaid.
The first context is the evolution of China's "Three Warfares" doctrine. In 2003, the Chinese People's Liberation Army formally designated "public opinion warfare, psychological warfare, and legal warfare" as integral components of military operations. This institutionalized the recognition that war is fought simultaneously not only in the physical battlefield but also in the information and legal spaces. Initially, influence operations primarily relied on traditional media such as newspapers, television, and radio, but the explosive spread of SNS in the 2010s fundamentally changed the operational environment.
The second context is China's AI industrial policy and military-civil fusion. In the 2017 "Next Generation AI Development Plan," China declared AI hegemony a national goal, aiming to become a global leader in core AI technologies by 2025. Crucially, this plan explicitly included "military-civil fusion," leveraging private sector AI technology for military and security purposes. Natural language processing technologies developed by tech giants like Baidu, Alibaba, and Tencent could be directly repurposed as content generation engines for influence operations. Since 2023, the rapid improvement in the performance of Chinese LLMs (e.g., Baichuan, Qwen, ChatGLM) has made high-quality, automated content generation in multiple languages, including Japanese, technically feasible.
The third context is the structural vulnerability of Japan's platform defense. Japan is one of the advanced democracies with the most delayed legal frameworks for disinformation countermeasures. The EU implemented the DSA in 2022, obligating platform operators to conduct systemic risk assessments and transparency reporting. Australia enacted the Online Safety Act in 2021, imposing a duty to promptly remove harmful content. Canada is also advancing the legislative process for its Online Harms Act. In contrast, Japan's response primarily centers on promoting "voluntary initiatives" by the Ministry of Internal Affairs and Communications' "Study Group on Platform Services," with virtually no legally binding obligations.
These three currents converged in 2025. China acquired AI technology, fostered through military-civil fusion, as the executive force for its "Three Warfares" doctrine. And Japan finds itself institutionally defenseless against this new threat.
Why "now"? There are three direct triggers. First, tensions in the Taiwan Strait have further escalated since 2024, leading to a sharp increase in China's need to "erode Japan's will to intervene in a Taiwan contingency." Second, the extensive use of AI-generated content in the 2024 U.S. presidential election completed the "field test" of AI influence operations, confirming their technical maturity. Third, the significant decline in moderation capabilities after the acquisition of X (formerly Twitter) dramatically lowered the "barrier to entry" for attackers.
Furthermore, the change in Japan's political situation since the beginning of 2025 cannot be overlooked. Following the inauguration of Prime Minister Sanae Takaichi, Japan's foreign and security policies are becoming more explicitly China-focused. The acceleration towards achieving 2% of GDP for defense spending, strengthening defense capabilities in the Southwest Islands with a Taiwan contingency in mind, and the full-scale operation of the Economic Security Promotion Act—all of these are factors that strengthen China's motivation to "sway Japanese public opinion through information warfare."
Historically, technological innovation in information warfare has always disadvantaged the defense. The invention of the printing press enabled the propaganda wars of the Reformation, radio enabled Nazi mass manipulation, and SNS enabled Russia's intervention in the 2016 U.S. presidential election. AI is a technology that further exacerbates this "attacker's advantage" structure. This is because AI dramatically lowers the cost of attack while not lowering the cost of defense. If it costs $0.01 to generate one piece of false information, it costs over $100 to detect, verify, and refute it. This asymmetry is the inherent threat of AI public opinion weapons and why traditional "fact-checking" and "media literacy education" cannot fundamentally address it.
The delta: The shift in China's information warfare against Japan from "human-intensive tactics" to "AI automation" has fundamentally changed the scale, speed, and cost structure of attacks. Traditional defensive measures (fact-checking, media literacy education, platform self-regulation) are structurally unable to keep pace with the scale of this new threat. Japan is one of the advanced democracies with the most delayed legal frameworks for disinformation countermeasures, and this institutional void provides a strategic attack surface for "AI public opinion weapons."
🔍 Between the Lines — What the News Isn't Saying
The core truth the news isn't telling is that China's AI public opinion weapons are not solely aimed at "attacking Japan." The true target is the effectiveness of the Japan-U.S. alliance in a Taiwan contingency, and manipulating Japanese public opinion is merely "preparatory groundwork" for that. The reason the Ministry of Defense dedicated a separate chapter to cognitive domain threats in its white paper is likely due to an accumulation of concrete AI influence operation detection cases, which cannot be publicly disclosed for intelligence reasons. Another unspoken truth is that Japan's delay in platform regulation, under the guise of "consideration for freedom of expression," is actually influenced by lobbying efforts against tech industry regulation.
NOW PATTERN
Narrative Hegemony × Platform Dominance × Institutional Decay
China's AI public opinion weapons are offensive tools designed to seize "narrative hegemony," utilizing the "platform dominance" structure of SNS platforms as an attack vector, and Japan's "institutional decay" (or more accurately, lack of institutional development) in disinformation countermeasures provides a defensive vulnerability, thus functioning in a three-layered structure.
Intersection of Dynamics
These three dynamics form a "vicious cycle triangle" that amplifies each other. China's AI public opinion weapons generate a massive volume of narratives to seize "narrative hegemony." These narratives are distributed on "platforms" with diminished moderation capabilities. And Japan lacks the "institutions" to regulate these platforms.
The most dangerous aspect of this three-layered structure is that **solving any one part will not solve the whole.** Even if Japan were to introduce platform regulations comparable to the EU (addressing "institutional decay"), if AI-generated content detection technology cannot keep up, the flood of narratives will not stop (leaving "narrative hegemony" unresolved). Even if AI detection technology dramatically improves, if platform operators are not obligated to implement it (leaving "platform dominance" unresolved), the technology will go to waste.
What is even more serious is the structure where time works to the advantage of the attacker. AI technology is evolving day by day; the quality of generated content continues to improve, and detection becomes increasingly difficult. Meanwhile, establishing legal frameworks takes years, and developing international standards requires even longer. This "gap between the speed of technological evolution and the speed of institutional development" functions as the greatest structural disadvantage for the defense.
The higher the tensions in the Taiwan Strait, the greater China's incentive to "manipulate Japanese public opinion." And the longer it takes for Japan to implement countermeasures, the longer China can continue to build "beachheads" in the information space. This temporal asymmetry is the most crucial structural factor linking the three dynamics, and it is why the problem of "AI public opinion weapons" is not merely a technological issue but a **national security issue.**
📚 Pattern History
1938-1945: Nazi Germany's Radio Propaganda (Volksempfänger)
Each time a new media technology emerges, authoritarian regimes repurpose it for propaganda before democratic nations do.
Structural similarities with this case: Nazi Germany was the first to extensively use the new technology of radio for propaganda. They mass-produced inexpensive receivers, "Volksempfänger" (People's Receiver), and dominated public perception through one-way information dissemination. Democratic nations were always reactive. AI public opinion weapons are following the same pattern.
2014-2016: Russia's IRA (Internet Research Agency) Intervention in the U.S. Presidential Election
Foreign election interference exploiting the structural vulnerabilities of SNS platforms takes several years to detect, and several more years to build effective countermeasures after detection.
Structural similarities with this case: Russia's IRA began its activities in 2014, but American society recognized the threat only after the 2016 presidential election. It took several more years to reach Facebook's congressional testimony, discussions on the Foreign Agents Registration Act, and platform regulation debates. Japan may currently be in the "pre-detection" phase of this timeline.
2019-Present: China's Information Warfare Against Taiwan (Systematization of "Cognitive Warfare")
Information warfare tactics field-tested in Taiwan are horizontally deployed to neighboring countries (Japan, Philippines, South Korea).
Structural similarities with this case: Taiwan faced China's SNS influence operations since the 2019 elections and by 2022 had built a systematic defense posture as "cognitive warfare." Led by the Digital Minister (then Audrey Tang), they deployed a strategy of "countering lies with humor." Japan can learn many lessons from Taiwan's experience, but Taiwan is a Chinese-speaking region, so its linguistic proximity differs from Japan's. AI translation bridges this gap.
2022-2023: Full-Scale Information Warfare in the Ukraine War
Physical military actions and information warfare unfold simultaneously, with "preemptive strikes" in the information space preceding physical actions.
Structural similarities with this case: In Russia's invasion of Ukraine, a large-scale disinformation campaign was launched weeks before military action. The false narrative that "Ukraine was planning genocide" served to "justify" military action. If China is preemptively deploying information warfare against Japan with a Taiwan contingency in mind, it is an application of the same pattern.
2024: The Year AI-Generated Content Was First Deployed on a Large Scale in Major Elections
The democratization of AI (lower cost, generalization) dramatically lowered the barrier to entry for influence operations, spreading them beyond state actors.
Structural similarities with this case: In the 2024 U.S. presidential election, Indian general election, and EU parliamentary elections, large volumes of AI-generated deepfake videos and fake news articles circulated. OpenAI reported disrupting multiple election-related influence operations. This experience serves as a technical template for AI information warfare against Japan from 2025 onwards.
Patterns Revealed by History
The pattern consistently shown by these five historical precedents is clear: **Each time a new communication technology emerges, authoritarian regimes repurpose that technology as a propaganda weapon before democratic nations do, and the response of democratic nations is consistently delayed by several years.**
With radio, the Nazis took the lead, and democratic nations did not implement effective countermeasures until the start of World War II. With SNS, Russia's IRA had been active since 2014, but the U.S. recognized the threat in 2017. Taiwan faced China's information warfare in the 2019 elections and built its defense posture over three years. In the Ukraine War, information warfare preceded military action. And in 2024, AI was deployed in practice as a tool for election interference.
What this history teaches is that **if Japan is currently in the "detection and recognition" phase, it is highly likely to take another 2-3 years to build effective countermeasures.** And those 2-3 years are also the 2-3 years during which tensions in the Taiwan Strait will further escalate. The pattern of history strongly suggests the importance of the defense acting before it's "too late." Taiwan is the only successful example of a democratic nation proactively building a defense posture, and how much Japan can learn from the Taiwan model will be the watershed moment going forward.
🔮 Next Scenarios
**Gradual institutional development, but unable to keep pace with AI evolution.** The Japanese government will submit some form of disinformation countermeasures bill to the Diet by 2026, but it will not be a comprehensive regulation like the EU's DSA, remaining limited to transparency obligations (e.g., reporting obligations for influence operations by platform operators). The Ministry of Internal Affairs and Communications-led "cooperation framework with platform operators" will be expanded, and an information-sharing mechanism with major platform operators will be established, but its legal binding force will be weak. The Ministry of Defense will increase the budget for "cognitive domain" countermeasures and proceed with the establishment of specialized units, but the full-scale operation of these organizations will take 2-3 years. Information sharing with Taiwan's Digital Ministry will be strengthened, but it will not reach the deep integration seen in the Five Eyes alliance. On the platform side, Meta and Google will continue to improve AI detection tools, while X's moderation system will not improve. Discussions regarding TikTok will proceed while observing U.S. trends, but no concrete regulations will be reached. In this scenario, China's AI influence operations will continue in an intermediate state where they "cannot be completely stopped, but the most blatant large-scale operations will be detected and exposed." The noise level in Japan's information space will gradually rise, but it will not lead to a societal crisis. However, in the event of a Taiwan contingency, there is a risk that this "intermediate state" of defense capability will be insufficient to prevent public opinion manipulation.
Implications for Investment/Action: Ministry of Internal Affairs and Communications' bill submission schedule, scale of Ministry of Defense's cognitive warfare budget, concrete progress in Japan-Taiwan information sharing, changes in content of major platforms' transparency reports
**Japan builds an advanced AI defense model.** Some shocking incident—for example, a "Japan-version IRA incident" where China's AI influence operations are massively exposed, or significant social disruption caused by deepfake videos—occurs, rapidly heightening a sense of crisis in Japanese society. This serves as a catalyst for accelerating bipartisan legislation on disinformation countermeasures. Under Prime Minister Takaichi's leadership, an "Information Space Defense Strategy" will be formulated in cooperation between the Digital Agency and the Ministry of Defense. A unique legal framework will be established, referencing the EU's DSA while considering Japan's constitutional constraints (freedom of expression). Transparency obligations for platform operators and mandatory disclosure of AI-generated content will be legislated. Furthermore, the development of AI detection technology, leveraging Japan's high technological capabilities, will accelerate, and a public-private AI fact-checking platform will become operational. A "Cognitive Warfare Defense Partnership" with Taiwan will be formally launched, enabling real-time threat information sharing. The establishment of an "Asia-Pacific Cognitive Warfare Countermeasures Center" in Tokyo, equivalent to NATO's StratCom Centre, will be decided. In this scenario, Japan will build "a cutting-edge system that reflects the latest knowledge precisely because it was late." However, the probability of this happening is low. Historically, there are few examples of democratic nations implementing this level of institutional reform "before an incident occurs."
Implications for Investment/Action: Occurrence of a large-scale AI influence operation exposure incident, bipartisan bill submission, movement towards organizational integration of the Digital Agency and Ministry of Defense, official announcement of Japan-Taiwan cognitive warfare cooperation
**Institutional development fails to keep pace, and AI public opinion weapons effectively pollute Japan's information space.** Japan's legal reforms are significantly delayed due to inter-ministerial coordination and political resistance, pushed back to 2027 or later. Meanwhile, China's AI influence operations become even more technologically sophisticated, acquiring not only the ability to generate natural Japanese text but also real-time audio and video. X's moderation system does not improve; instead, it is further reduced. TikTok's market share in Japan continues to expand, becoming a primary channel for information consumption among younger demographics, but ByteDance's algorithm transparency is not ensured. "Voluntary initiatives" with platform operators become hollow and have no substantial effect. As tensions in the Taiwan Strait further escalate towards 2027, China launches a large-scale information operation to guide Japanese public opinion towards "non-intervention in a Taiwan contingency." Narratives precisely targeted along existing fault lines in Japanese society, such as distrust of the Japan-U.S. alliance, U.S. military base issues, nuclear armament debates, and historical issues, are disseminated. In the worst-case scenario, Japanese public opinion splits during an actual Taiwan contingency, making rapid government decision-making difficult. The most serious consequence of information warfare—AI public opinion weapons affecting the physical military balance—becomes a reality. The probability of this scenario is 30%—not low. This is because historically, there are countless examples of democratic nations being "too late" in responding to information warfare.
Implications for Investment/Action: Delay in bill submission, announcement of moderation cuts by platform operators, rapid escalation of military tensions in the Taiwan Strait, occurrence of large-scale distribution of Japanese deepfakes
Key Triggers to Watch
- Outcome of the Active Cyber Defense Bill's Diet deliberation and the presence/absence of disinformation countermeasures clauses: By June 2026 (within the ordinary Diet session)
- Announcement of concrete legislation for the Ministry of Internal Affairs and Communications' "Institutional Response to False and Misleading Information": By September 2026
- Large-scale exposure incident regarding China's AI influence operations (public disclosure by OpenAI/Meta/Japanese government): Within 2026 (unpredictable, but high probability following U.S. examples)
- Significant military tensions in the Taiwan Strait (large-scale military exercises, maritime clashes, etc.): 2026-2027
- Significant change in X (formerly Twitter)'s content moderation policy: Within 2026
🔄 Tracking Loop
Next Trigger: Diet deliberation of the Active Cyber Defense Bill — Whether disinformation countermeasures clauses are included in the ordinary Diet session from March to June 2026 will be the first watershed moment determining the speed of Japan's institutional development.
Continuation of this Pattern: Tracking Theme: Building Japan's AI Disinformation Defense Posture — The next milestone is legislative debate in the 2026 ordinary Diet session, followed by concrete budget allocation for cognitive warfare countermeasures in the 2027 Defense White Paper.
>How do you read this? Participate in Prediction →
