The soldiers in your feed may just be artificial intelligence

Introduction: Hybrid warfare enters a new digital era

In 2026, the boundaries between reality and digital fiction have become thinner than ever. Generative artificial intelligence has evolved into a sophisticated propaganda tool capable of creating fake military identities, generating hyper-realistic video content, and manipulating public opinion on a global scale. The soldiers you see in your social media feed may not actually exist – they may be the product of advanced image and video generation algorithms specifically trained to mislead the civilian population and influence perceptions of armed conflict. This is no longer science-fiction speculation, but a documented reality by cybersecurity researchers, investigative journalists, and government agencies around the world.

phenomenon military deepfake represents one of the most dangerous applications of generative AI in the context of modern information warfare. Technologies such as GAN (Generative Adversarial Networks) and diffusion models (Diffusion Models) now allow the creation of high-resolution images and videos of soldiers, military vehicles, and combat scenarios that never existed before. The cost of producing this type of content has dropped dramatically, and the accessibility of the necessary tools has increased exponentially, opening the door for state and non-state actors alike.

How AI-based disinformation factories work

The technology behind virtual soldiers

At the heart of AI-based military disinformation operations is a series of converging technologies that, combined, produce content that is almost indistinguishable from reality. Text-to-image models State-of-the-art models, such as Stable Diffusion, Midjourney or DALL-E, can generate 4K resolution images of soldiers equipped with real weapons, in specific geographical locations, with correct lighting conditions and tactical details in a few seconds. Moreover, the models text-to-video adds an additional level of realism, allowing the creation of video sequences in which these virtual soldiers move, communicate and act in fabricated combat scenarios.

Another critical tool in the disinformers' arsenal is voice cloning technology – voice cloning technology. By analyzing just a few seconds of audio sample, AI models can accurately replicate the voice of a real general, government official, or military spokesperson, generating fake statements that sound authentic. The combination of deepfake video and voice cloning creates what experts call synthetic media attacks – synthetic media attacks – a new category of cyber threats that tests the limits of automated detection and human reason.

Automated amplification and distribution networks

Generating fake content is just the first step. AI-powered botnets is the distribution mechanism that turns a single piece of misinformation into a viral phenomenon. These networks of fake accounts, automatically managed by algorithms like LLM (Large Language Model), can simulate human behavior online, interacting with real content, commenting, distributing and socially validating false information. Advanced language models allow these bots to generate relevant contextual comments, answer questions and maintain coherent conversations over long periods of time, without human intervention.

Researchers from institutions such as Stanford Internet Observatory si EU DisinfoLab have documented coordinated campaigns in which thousands of fake accounts simultaneously shared AI-generated images of fabricated soldiers and civilian casualties, influencing public perception of active conflicts. The speed of these campaigns is far beyond the capacity of human fact-checking teams, making it the latency of misinformation – the interval between the publication of false content and its correction – must be long enough to cause significant reputational and social damage.

Documented cases from 2025-2026

Influence operations with artificially generated soldiers

The 2025 and 2026 reports confirmed the widespread use of synthetic military identities in the context of several regional conflicts. Intelligence agencies in European states have identified campaigns in which social media profiles with AI-generated soldiers were used to spread false narratives about loss of life, atrocities, or military successes. These profiles had detailed histories, coherent photos generated by AI models, and networks of friends artificially built over months before being activated for disinformation purposes.

A particularly worrying technical aspect is the use of consistent identity generation – the ability of AI models to generate dozens or hundreds of images of the same fictional character in different contexts, with sufficient visual coherence to convince the human eye. This technique, available through affordable commercial tools, allows the creation of digital legends – false identities with a seemingly authentic and lasting online presence.

Impact on the perception of armed conflicts

The effects of AI-based military disinformation are not abstract – they translate into concrete public reactions, influenced political decisions, and amplified social polarization. Cognitive psychology studies show that first exposure to information, even if it is later denied, leaves a persistent neurological imprint – a phenomenon known as illusory truth effectIn the context of information wars, this effect is systematically exploited: even if an image of a fake soldier is identified and reported, millions of users have already seen it and have partially integrated it into their model of understanding the respective conflict.

More, verification fatigue (fatigue check) is a real and measurable phenomenon among social media users: repeated exposure to misinformation and the effort to validate or invalidate it causes a progressive reduction in motivation and critical capacity, leading to a passive acceptance of information, regardless of its source.

Detection tools and technological countermeasures

AI vs. AI: The Battle for Digital Authenticity

The technological response to the proliferation of AI-based military disinformation involves the development of increasingly sophisticated detection systems. Deepfake classification models analyzes subtle inconsistencies in skin textures, eye movement, compression artifacts, and lighting inconsistencies that betray the artificial origin of an image or video. Major tech companies, such as Microsoft, Google and Meta, have invested heavily in such systems, integrating them directly into their platforms to detect and limit the distribution of synthetic content.

A technology with high potential in this context is C2PA (Coalition for Content Provenance and Authenticity) – an open technical standard that allows cryptographic metadata to be attached to media content at the time of its creation. By implementing C2PA, cameras, editing applications, and distribution platforms can form a verifiable digital chain of custody, allowing users to see provenance and history of changes of any media content. Adopting this industry-wide standard could revolutionize the ability to verify the authenticity of information in real time.

Invisible watermarking and fingerprinting

Another promising technical direction is invisible watermarking of AI-generated contentModels like SynthID Google DeepMind embeds an imperceptible digital signal into artificially generated images and audio, a signal that can later be detected by specialized tools, even if the content has been compressed, resized, or edited. While this technology is not foolproof—malicious actors can try to remove or alter the watermark—it adds an additional layer of traceability that increases the cost and complexity of disinformation operations.

At the institutional level, Defense Advanced Research Projects Agency (DARPA) invested in programs such as Media Forensics (MediFor) si Semantic Forensics (SemaFor), aimed at developing automated media authenticity analysis tools capable of processing large volumes of content in real time. These programs combine techniques computer vision, natural language processing and metadata analysis to build a complete profile of the analyzed media content and to estimate the probability of its synthetic origin.

Platform responsibility and legal framework

Emerging regulations in the AI ​​space and disinformation

In 2026, regulatory pressure on social media platforms and AI model providers increased significantly. AI European Act includes specific provisions regarding synthetic content with the potential for disinformation, imposing labeling and traceability obligations for AI applications classified as high-riskIn the United States, legislative debates regarding DEEPFAKES Accountability Act and other similar initiatives reflect the growing awareness of the threats posed by synthetic media in the context of national security.

Major platforms have implemented AI-generated content labeling policies, but their effectiveness remains limited in the absence of robust technical automatic detection mechanisms. The fundamental challenge consists of the asymmetry between the speed of synthetic content generation – which benefits from the computing power of modern GPUs and the accessibility of open-source models – and the speed of detection and moderation, which involves slower processes and limited human resources.

Media literacy and digital resilience

Beyond technological and legislative solutions, digital media education remains a key pillar of defense against AI-based military disinformation. Citizens’ ability to critically evaluate information sources, recognize signs of synthetic content, and utilize fact-checking tools is a line of defense that complements technical solutions. Organizations like First Draft, Bellingcat si EEAS East StratCom Task Force offers educational resources and practical tools for journalists, educators and ordinary users.

The concept of prebunking – pre-inoculation against disinformation through controlled exposure to manipulation techniques – has demonstrated effectiveness in clinical trials and is now integrated into national media education programs in several European countries. Combining prebunking with AI tools for automatic detection creates a multi-layered defense system, more resistant to the rapid evolution of disinformation techniques.

The future of information warfare: emerging scenarios and risks

As AI models continue to evolve, military disinformation generation capabilities will become even more sophisticated. Real-time video generation models, combined with real-time face swapping and with interactive digital avatars, they will allow the creation of fake live broadcasts – fake livestreams – with virtual soldiers or military officials answering questions in real time and interacting with the audience. This evolution will test not only technical detection systems, but also human perception as a whole.

Another emerging risk is represented by personalization of misinformation – using personal data and psychographic profiles to tailor synthetic content to each user’s specific cognitive vulnerabilities. AI models can analyze a user’s browsing history, content preferences, and emotional reactions to generate and deliver military disinformation calibrated to maximize psychological impact on a specific target. This capability transforms disinformation from a mass tool to one of surgical precision.

In this context, international cooperation between governments, academic institutions, tech companies and civil society is becoming not just desirable, but imperative. Common technical standards, sharing of threat intelligence and coordinated investments in research and education are the pillars of an effective response to the challenges of AI-based military disinformation in the decade ahead.

Conclusion: Vigilance in the era of virtual soldiers

The soldiers in your feed may, indeed, just be artificial intelligence. Modern information warfare has long since transcended the barriers of manipulated text and photography, entering the era of high-fidelity synthetic media, produced on an industrial scale by AI models accessible to anyone. Understanding the technical mechanisms underlying these threats, adopting available verification tools, and cultivating an active critical spirit in information consumption is not only a competitive advantage, but a necessity for any informed citizen of the digital age.

The answer to this threat is not to give up consuming online information, but the sophistication of their evaluation process – an essential digital competency that must be cultivated systematically, from the individual to the institutional and governmental level.

You have certainly understood what is new in 2026 related to artificial intelligence. If you are interested in deepening your knowledge in the field, we invite you to explore our range of courses structured by roles and categories in AI HUBWhether you're just starting out or want to brush up on your skills, we have a course for you.