This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
Russia has extensively utilized AI-driven tools, including bot networks and generative models, to conduct large-scale disinformation campaigns aimed at eroding Western support for Ukraine. These operations exploit social media platforms to spread propaganda, manipulate public opinion, and create divisive narratives. While AI is also being explored as a tool to combat this disinformation, the scale and speed of AI-generated content pose significant challenges to platform moderation and information security.
The Russian government has been stepping up its foreign influence campaigns by using artificial intelligence and emerging technologies to spread disinformation and sow distrust in policies supportive of Ukraine. Part of the strategy includes carrying out influence laundering operations by disseminating their messages to the American public via allies inside nominally independent organizations, according to a recent declassified analysis.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Using AI systems to conduct large-scale disinformation campaigns, malicious surveillance, or targeted and sophisticated automated censorship and propaganda, with the aim of manipulating political processes, public opinion, and behavior.
Also classified as
Coordinated networks of fake accounts directed online harassment and intimidation at citizens expressing support for Ukraine.
The massive influx of automated bot activity and synthetic media pollutes social media platforms, degrading the shared information ecosystem.
The incident involves coordinated, state-sponsored campaigns using AI-driven bot networks and deepfakes to conduct large-scale disinformation and manipulate public opinion regarding the war in Ukraine.
Human
Due to a decision or action made by humans
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The disinformation campaign and deployment of bot networks were intentionally directed by human actors (the Russian government and intelligence services) to influence public opinion.
Added categories peak at 4: Severe (Epistemic Harm, Inferred)
Russia conducted large-scale, state-sponsored information operations using AI-driven bot networks, deepfakes, and automated amplification to manipulate public opinion and erode Western support for Ukraine. The incident represents a severe information warfare threat and a substantial challenge to democratic sovereignty and societal stability.
The incident involves coordinated networks of thousands of automated bot accounts (swarms) acting in unison to amplify specific narratives, like posts, and manipulate trending algorithms. Report 2 notes 'massive bot networks' and '870 accounts had a score of one' on Botometer, demonstrating collective orchestrated behavior.
Unacceptable Risk: The reports describe the use of AI-driven bot networks and deepfakes for large-scale behavioral influence and cognitive manipulation of public opinion, which poses significant dangers to society and democratic processes.
The primary goal of the AI systems (bots and generative models) was to generate and amplify social media content to spread disinformation, while platform recommendation algorithms further curated and promoted this content to users.
The reports provide extensive, multi-perspective coverage of the disinformation campaigns, including academic research using Botometer, official statements from tech platforms like Twitter, and intelligence assessments from U.S. officials. The role of AI in generating deepfakes and coordinating bot networks is clearly documented with specific metrics (e.g., 75,000 accounts removed, Botometer scores).
Some platforms, like Facebook in Report 1, argued that certain flagged accounts were operated by single human users rather than coordinated bot networks, suggesting some platform manipulation may be organic rather than automated.
Technology
AI-Enabled Bot Networks
Automated bot accounts rapidly amplify Russian disinformation and hashtags.
Generative AI and Deepfakes
Generative tools create realistic fake videos and texts at near-zero cost.
Recommendation Algorithms
Platform algorithms prioritize viral content over factual accuracy.
Data Inputs
Open Platform APIs
Flexible APIs allow bad actors to automate accounts and scrape data easily.
Biased Training Data
LLMs trained on Western data contradict Russian state-approved narratives.
Lack of Proactive Detection Data
AI detection models struggle with novel disinformation without human labeling.
Human Factors
High Susceptibility to Deepfakes
Users trust manipulated media like fabricated videos of political leaders.
User Polarization
Polarized audiences readily accept and amplify unverified conspiracy theories.
Mass Flight of IT Talent
Brain drain of Russian tech experts limits domestic AI safety development.
Process and Methods
Inadequate Automated Moderation
Platforms fail to proactively identify and remove automated bot networks.
Government Account Loophole
Rules restricting state media do not apply to official embassy accounts.
Fragmented Platform Policies
Inconsistent moderation across platforms creates safe havens for propaganda.
Management
Dismantling Safety Teams
Platforms like X dismantled teams hunting for election interference.
Prioritizing Engagement over Safety
Platform algorithms prioritize viral content over factual accuracy.
Inadequate Risk Assessment
Management failed to anticipate the scale of bot networks using open APIs.
Regulatory Environment
Fragmented Global Regulations
Divergent AI and data laws make coordinated global moderation difficult.
Punitive Disinformation Laws
Strict domestic laws force Russian AI models to self-censor on key events.
Western Tech Sanctions
Restrictions on hardware limit the computing power needed for safe AI models.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
