State-Backed AI Propaganda Operations Expand Production Capabilities
State-linked operators from Iran, Russia, and China are actively utilizing frontier artificial intelligence models, including Anthropic’s Claude and OpenAI’s ChatGPT, to automate propaganda production, draft internal doctrines, and orchestrate influence campaigns. According to threat reports released by frontier AI companies and social media platforms, these foreign actors are successfully increasing their operational capacity and output speed, though evidence remains scarce that these automated efforts are actually shifting public opinion.
Iran and Russia use AI for state propaganda
Iranian state propagandists used Claude to draft doctrines for influence campaigns and to plan the Supreme Leader’s funeral, according to disclosures from Anthropic. In addition, an Iranian opposition group, the MEK, ran a shared AI agent equipped with memory files containing banned words and detection-evasion rules to sustain automated content production.
Russian state media actors have similarly integrated large language models into their daily workflows. A former Sputnik Moldova editor-in-chief used AI to churn out articles and push fabricated claims ahead of Moldova’s 2025 election, while a staffer at a Russian state media outlet relied on AI to write on-air chyrons.
Chinese Influence Campaigns Targeting American Debates
China-linked operators have deployed American artificial intelligence to meddle in domestic American debates over AI infrastructure and data centers. In June, OpenAI reported that accounts likely operated by a Chinese tech contractor falsely claimed that ChatGPT had leaked user data. Another cluster of accounts used ChatGPT to generate political cartoons blaming data centers for rising electricity bills among Americans.
Administrative Transformation and Autonomous Cyber Operations
The primary shift detailed in recent threat reports is not merely content generation, but administrative and operational transformation. AI models are taking over back-office tasks such as audience research, timing strategies, data processing, and control software automation. Chinese operators utilized ChatGPT to write code automating logins across multiple social media platforms.
Distribution bottlenecks limit the reach of AI propaganda
While generative and agentic AI have drastically reduced the overhead and increased the scale of propaganda creation, distribution remains a distinct bottleneck. Anthropic noted that a French ad agency network published roughly 9,000 articles across 70 fake news sites in 20 languages that drew almost no real engagement. Similarly, OpenAI rated its China-linked campaigns as “Category One,” indicating they failed to spread beyond their own controlled accounts.
Because capturing organic attention and earning trust is difficult, state-linked actors increasingly attempt to bypass the challenge by utilizing established influencers or co-opting real accounts. However, the core challenge for foreign influence networks remains unchanged: while AI scales production capacity effortlessly, manufacturing trusted distribution and measurable changes in public opinion continues to depend on human audiences and traditional media amplification.
AI models help draft propaganda but rarely change opinion
What specific AI models were identified in recent threat reports regarding foreign influence?
Threat disclosures from Anthropic and OpenAI identified operators utilizing models such as Claude and ChatGPT to draft propaganda, write code, manage administrative workflows, and generate political content.
Did AI-generated influence campaigns successfully change public opinion?
According to safety and threat reports from AI companies and platforms, while AI significantly increases production capacity, speed, and persistence, there is little evidence that these automated operations have actually succeeded in altering public opinion.
How are state-backed actors utilizing AI beyond simple text generation?
Operators are using AI agents for administrative tasks, audience research, workflow automation, and writing code to bypass platform logins.
Worth a look