The idea that certain individuals on the street might not be acting autonomously — that their behavior, location, and interactions are coordinated by an external system — sounds paranoid until you examine the documented reality of state-sponsored intelligence operations. Multiple governments operate large-scale human intelligence networks in which individuals are deployed to specific locations, given specific behavioral scripts, and tasked with surveillance, influence, or disruption objectives that they may not fully understand. These operations are not speculative. They are documented in declassified intelligence assessments, federal court proceedings, congressional testimony, and the published doctrines of nation-state military and intelligence services.
China’s Sharp Eyes program (Xueliang Gongcheng) represents perhaps the most comprehensive civilian surveillance and reporting network ever built. Launched as a national initiative in 2015 and integrated into China’s 13th Five-Year Plan, Sharp Eyes extends the country’s existing video surveillance infrastructure by connecting cameras in rural villages and urban neighborhoods to centralized monitoring platforms that are accessible to local Communist Party officials and designated civilian monitors. The system is explicitly designed to enlist ordinary citizens in surveillance — residents can watch camera feeds on their televisions or smartphones and report suspicious activity to local authorities. By 2020, the program had been deployed across hundreds of cities and counties. The underlying camera network feeds into China’s broader Skynet and Safe Cities surveillance systems, which the Carnegie Endowment for International Peace has documented as comprising over 600 million cameras nationwide, many equipped with AI-powered facial recognition supplied by companies including Hikvision, Dahua Technology, SenseTime, and Megvii. The Sharp Eyes system explicitly blurs the line between civilian and intelligence operative — any participant with access to the monitoring feeds becomes a node in the state surveillance apparatus.
Coordinated inauthentic behavior online — the use of fake or manipulated accounts to simulate grassroots activity — has been documented at industrial scale by every major social media platform. Meta’s quarterly Adversarial Threat Reports have identified and removed coordinated networks originating from Russia, China, Iran, Bangladesh, and dozens of other countries. Between 2017 and 2024, Meta removed over 200 coordinated inauthentic behavior networks comprising millions of accounts. The Internet Research Agency (IRA), based in St. Petersburg, Russia, employed over 1,000 people by 2018 and operated 24 hours a day creating content designed to appear as authentic American political speech. The IRA’s operations were detailed in the Mueller investigation’s indictment of 13 Russian nationals in February 2018, which documented specific budgets, management structures, travel to the United States for reconnaissance, and the creation of entire fictitious American activist organizations — including accounts that organized real-world rallies attended by real Americans who had no idea the event was conceived in a Russian office building.
China’s equivalent online operation, documented by researchers at Harvard’s Shorenstein Center and Oxford’s Computational Propaganda Project, is known informally as the 50 Cent Army (wumao). A 2017 study by Harvard political scientist Gary King estimated that the Chinese government fabricates approximately 448 million social media posts per year on domestic platforms, authored by government employees — not paid freelancers — who post as part of their official duties. These posts are strategically timed to coincide with events that might generate public dissatisfaction and are designed not to argue with critics but to distract and redirect public attention. The scale of this operation means that a significant percentage of the political discussion visible to Chinese internet users on any given day is synthetic — authored by state operatives following centralized directives.
The physical-world equivalent of coordinated inauthentic behavior is well-documented in the intelligence literature. The FBI’s COINTELPRO program, which operated from 1956 to 1971, deployed agents and informants who infiltrated domestic organizations, attended meetings, organized events, and built relationships — all while reporting to and being directed by their FBI handlers. The Church Committee hearings in 1975 exposed the scope of these operations, which targeted civil rights organizations, antiwar groups, and political movements across the ideological spectrum. Declassified documents show that COINTELPRO operatives were given specific behavioral instructions: who to approach, what positions to advocate, which organizational disputes to inflame. The individuals in these networks appeared to be autonomous citizens exercising their own judgment. They were, in operational fact, executing coordinated scripts.
Modern military doctrine has formalized the integration of AI into these operations. The People’s Liberation Army’s concept of Intelligentized Warfare, articulated in its 2019 defense white paper, explicitly calls for the use of artificial intelligence to coordinate military operations, intelligence collection, and information warfare at machine speed. The U.S. Department of Defense’s Joint All-Domain Command and Control (JADC2) initiative, with a projected investment exceeding $30 billion, aims to connect every sensor, every shooter, and every command node across all five warfighting domains — land, air, sea, space, and cyber — through AI-enabled decision networks. Project Maven, initiated by the Pentagon in 2017, was the DOD’s first operational AI program and used machine learning algorithms developed initially by Google (which withdrew from the contract after employee protests) to analyze drone surveillance footage and identify targets. The program was later expanded and absorbed into the National Geospatial-Intelligence Agency’s operations.
Autonomous decision systems in military contexts have advanced beyond target identification into the realm of engagement recommendations. The Israeli Defense Forces’ use of AI targeting systems in Gaza, reported by publications including +972 Magazine and The Guardian, included a system internally called Lavender that allegedly generated target lists of suspected militants using machine learning trained on surveillance data, communications intercepts, and social network analysis. According to these reports, human review of the system’s targeting recommendations was minimal — sometimes as little as 20 seconds per target. Whether or not the specific details of these reports are fully confirmed, they reflect a documented trajectory in military AI: the progressive automation of decisions that were previously reserved for human judgment, driven by the volume of data that exceeds human analytical capacity and the speed at which modern conflicts require response.
DARPA’s Social Media in Strategic Communication (SMISC) program, funded from 2011 to 2015, explicitly researched methods for detecting and countering influence operations on social media — and, necessarily, developed a detailed understanding of how to conduct them. The program’s stated objectives included identifying the formation, development, and spread of ideas and concepts on social media, recognizing persuasion campaign structures, and modeling the dynamics of how narratives spread across networks. A separate DARPA program, the Information Innovation Office’s Active Social Engineering Defense (ASED), focused on detecting social engineering attacks. The dual-use nature of this research — understanding how influence operations work requires understanding how to build them — is a recurring pattern in defense research that applies equally to cybersecurity, biological defense, and electronic warfare.
The convergence of physical surveillance networks, AI-driven behavioral coordination, and industrialized online influence operations creates an environment where the boundary between authentic and directed behavior becomes genuinely difficult to determine. When state intelligence services can deploy human operatives with scripted behaviors, amplify their messaging through millions of synthetic online accounts, monitor the results through comprehensive camera and signals intelligence networks, and adjust their tactics in real time using AI analytics — the resulting system functions as a distributed intelligence organism. Individual nodes in this system may or may not be aware of their role. The Chinese Sharp Eyes participant watching a camera feed on their phone, the IRA employee crafting Facebook posts in St. Petersburg, the FBI informant attending a political meeting in 1968 — each is performing a specific function within a coordinated system, and none has visibility into the full architecture of the operation they serve.
This documented reality — that multiple nation-states operate coordinated networks of human and artificial agents designed to surveil, influence, and control civilian populations — is the factual foundation for the investigations published on this site. The programs described above are not theories. They are budgeted initiatives with named program managers, published doctrine, peer-reviewed analysis, and in many cases, congressional oversight (or documented lack thereof). When this site examines the intersection of surveillance technology, military AI, underground infrastructure, and advanced biotechnology, it does so against the backdrop of these verified programs. The question is never whether governments coordinate inauthentic behavior at scale. They demonstrably do. The question is what capabilities exist at classification levels above what has been made public, and what those capabilities mean for the autonomy and security of individuals who are not party to the programs being run around them.