AI Can Be Used for Cognitive Warfare

China's state security minister Chen Yixin publicly warned that hostile forces are abusing generative AI technologies such as deepfake audio and video, AI-generated text and images, and "intelligent troll armies" to conduct cognitive warfare against the country, calling it a direct threat to political, institutional and ideological security. The warning reflects Beijing's concern that AI-powered repetition at scale, exploiting the psychological illusory truth effect, can erode public skepticism and protect those in power from criticism.

The People’s Republic of China PRC has publicly reflected on its fear that hostile forces can use AI https://archive.is/NNk2J for ‘cognitive warfare’. Previous concerns in the PRC relate to the idea that AI can constitute a rival power faction within the country, threatening the Chinese Communist Party’s hold on power. https://www.nytimes.com/2026/09/14/world/asia/china-ai-security-risks-anthropic.html The PRC has admitted that it fears AI-powered cognitive warfare for the same reason it fears artificial general intelligence AGI , both can potentially threaten its monopoly over Chinese society: “Hostile forces and people with ulterior motives are abusing generative AI technologies such as deepfake audio and video, AI-generated text and images, and ‘intelligent troll armies’ to cheaply and on a mass scale fabricate political rumours, disseminate harmful information and incite divisive sentiments,” said state security minister Chen Yixin https://archive.is/NNk2J . “These activities constitute an attack on public opinion and cognitive warfare against China, directly threatening our political security, institutional security and ideological security”. AI so far has had limited use in actually fooling people. The fear used to be that generative AI could create deep fakes of fake events, permanently compromising our ability to discuss verified, real events online. These fears remain unrealized, as even convincing fake imagery and video still lack the extra 1 or 2% of quality and credibility to fool people en masse. Generative AI multimedia is largely a storytelling tool https://thecarrierwave.substack.com/p/the-killer-use-case-for-ai-in-social for memes https://thecarrierwave.substack.com/p/kirkversary-a-bizarre-ai-carnival , jokes, and political attacks. We have speculated the killer use case of AI might just be to flood our online platforms with so much noise, real communication becomes impossible https://thecarrierwave.substack.com/p/the-information-wars-a-retrospective . However, there is another use case beyond distraction that AI is tailor-made for, and it involves exploiting a deep-seated flaw in human psychology. The “Repeater Bias”, otherwise known as illusory truth effect https://en.wikipedia.org/wiki/Illusory truth effect states that the more often you hear a statement, the more likely you are to believe that it is true. If you've ever wondered why advertisers will buy up several advertising blocks on the same TV program to show you the same advertisement multiple times in an hour, this is the reason. Repetition creates permanent relevance in your brain, subconsciously, without your consent. And there are no known countermeasures. This is also why politicians repeat statements over and over https://www.usatoday.com/story/news/politics/onpolitics/2017/02/16/mess-fake-news-disaster-trumps-repetition-advertising-tactic/98014444/ , even if they are absurd on the surface. This technique when applied with AI-powered bots at scale could produce a similar effect, without any regional limitations. Simply program robots to repeat something, no matter how absurd or outside of reality, and you will slowly start to move the needle of public opinion in the direction of your choosing. The science says https://pmc.ncbi.nlm.nih.gov/articles/PMC8116821/ it doesn't matter what the idea is, who the audience is, history, context, etc. Over time, public opinion might not significantly shift, but it will at the least soften to an idea if subjected to repeated exposure. This “opinion softening” is strategically important, as in many cases in information warfare, the goal here is not to convince a population of something new. The goal is simply to erode skepticism over time and utilize margins before elections if elections are allowed to occur to protect those in power from criticism and public opposition. AI is a tailor-made technology for this kind of cognitive warfare. Authoritarian states have been deploying this technique for years. Previously there was a hard cap on the number of times you could repeat something. You only had so many state-sponsored journalists. You only had so many submissive politicians. Now, realistically, there is no limit. You can program bots to push your own message to a near-infinite audience, 24 hours a day. What could be done to combat this new form of warfare? Not a whole lot if you insist on maintaining a presence on public social media. This technology will accelerate a trend we had already seen coming: the idea that public opinion expressed on social media cannot be trusted. Open networks such as X.com, while entertaining and still the de facto source for many kinds of breaking news and commentary, must be treated with a permanent skepticism. Even before the proliferation of uncontrolled bots, social media virality very rarely coincided with accuracy. Do not delude yourself into thinking that social media is anything but the junk food of your information diet. Any opinions seen in the comment section are anything but a concentrated effort to repeat specific narratives, either for propaganda or profit. It is human instinct to see what other people think and form your opinion based on that previous consensus. In trusting the opinions of others, we take cognitive shortcuts in order to understand a world too large to fully comprehend by ourselves. This instinct has unfortunately been hijacked by malicious actors seeking to poison our understanding of the world. There might not be a simple, effective solution, but as always awareness is the first step. The best countermeasure to being misled and misinformed is maintaining strict standards in the sources you choose to trust. The comment section should be no different.