This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
The Matryoshka disinformation campaign is a coordinated influence operation that uses AI-generated voice clones to impersonate academics and experts in doctored videos. These videos, which repurpose original footage to include false political narratives, aim to promote pro-Russian propaganda and undermine support for Ukraine. The campaign has been observed operating on platforms including X and Bluesky, utilizing bot networks to amplify the reach of the manipulated content.
The Matryoshka disinformation campaign allegedly used AI to impersonate academics, reportedly spreading claims supporting Russia and urging Ukraine's surrender. These videos are said to have misrepresented scholars’ views in order to amplify pro-Russian propaganda. The campaign reportedly exploited global social media platforms to allegedly mislead viewers and undermine trust in credible academic voices while fueling misinformation about the Ukraine conflict.
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
The campaign generated and spread fabricated claims and false narratives using cloned voices of trusted academics.
The Matryoshka campaign is a coordinated Russian influence operation using AI voice-cloning to spread political disinformation at scale across social networks like X and Bluesky.
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 incident is driven by human actors intentionally using AI voice cloning tools post-deployment to generate and spread disinformation.
Added categories peak at 4: Severe (Epistemic Harm, Inferred)
The Matryoshka campaign represents a coordinated Russian state-aligned disinformation operation that leverages AI voice cloning to impersonate trusted Western academics. By fabricating statements to undermine support for Ukraine, the campaign demonstrates an evolved, industrialized capability in state-sponsored information warfare, though its direct impact on core government operations remains limited.
The AI systems used in this incident are voice cloning and deepfake generation tools. They perform specific, well-defined functions of replicating audio based on direct inputs provided by human operators, without exhibiting autonomous planning or goal-directed behavior of their own.
Risk Level 3: Limited Risk. The report describes the use of AI-generated content such as deepfakes ('The video is a deepfake whose audio has been manipulated' and 'cloning the voices of real professors using artificial intelligence (AI) tools'), which falls under transparency obligations for AI-generated content under the Act.
The primary AI technology used was voice cloning to replicate the voices of academics, which was then used to generate social media propaganda videos.
The reports provide consistent, detailed accounts of the Matryoshka campaign across multiple sources (AFP, The Insider, Bot Blocker). The mechanism of the AI harm (voice cloning of specific academics to spread pro-Russian propaganda) is clearly documented and verified by independent institutions like the University of Bristol.
None. The evidence clearly points to a coordinated state-aligned disinformation campaign using synthetic media.
Technology
AI Voice Cloning
Cloned real academics' voices to make fake statements sound authentic.
Industrialized Deepfakes
Automated production scaled up the generation of fake voiceovers.
Data Inputs
Hijacked Real Videos
Original university videos were harvested to anchor fake audio.
Human Factors
Trust in Authority
Users trusted videos due to prestigious university associations.
Stolen User Accounts
Compromised profiles were used to spread and amplify deepfakes.
Process and Methods
Reactive Moderation
Platforms relied on user reports rather than proactive detection.
Coordinated Bot Networks
Bots systematically reposted deepfakes to manipulate algorithms.
Management
No Proactive Defenses
Management failed to deploy proactive automated detection tools.
Growth Over Security
Platform prioritized user migration over scaling security systems.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
