Presidency of Moldova Reportedly Refutes Purported Deepfake Video Slandering President Maia Sandu

A deepfake video reportedly depicting President Maia Sandu's image and voice was released in Moldova, allegedly portraying her in a negative light to sow division and undermine democratic institutions. This video reportedly appeared on Telegram and was linked to Russian disinformation efforts.

Risk classification

• Primary risk domain: 4 Malicious actors
• Primary risk subdomain: 4.1 Disinformation, surveillance, and influence at scale

The incident involves the coordinated use of AI-generated deepfakes by political actors to conduct a large-scale disinformation campaign aimed at manipulating democratic processes and public opinion in Moldova.

Additional risk subdomains

• 3.1 False or misleading information: The AI system was used to generate false and misleading video and audio content depicting the President making fabricated statements.
• 6.5 Governance failure: The report highlights that Moldova has not yet taken legislative steps or developed a national strategy to monitor or regulate the use of generative AI.

Causal factors

• Entity: Human
• Intent: Intentional
• Timing: Post-deployment

The deepfake videos were intentionally created and deployed by human malicious actors using post-deployment generative AI tools to manipulate public opinion.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk: The report describes the creation of AI-generated deepfakes, which fall under the category of AI systems posing a moderate risk that require specific transparency obligations under the EU AI Act.

AI system and alleged parties

• AI system: None named
• AI purpose: Deepfake Video Generation; Voice Generation
• Behaviour type: Tool
• Alleged developer: Unknown voice cloning technology developers, Unknown deepfake technology developers, Unknown deepfake creator, Russian propagandists
• Alleged deployer: Unknown deepfake creator, Russian propagandists, Pro-Kremlin actors
• Alleged harmed parties: Presidency of Moldova, National security and intelligence stakeholders, Maia Sandu, Government of Moldova, General public, Epistemic integrity, Democracy

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Minor. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Negligible
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Minor, indirect Minor
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Negligible, indirect Negligible
• Democracy: direct Minor, indirect Substantial
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Negligible
• Epistemic: direct Minor, indirect Substantial
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Malicious content

Reported: The report explicitly describes the creation and spread of malicious deepfake videos targeting the Moldovan President.

Directly caused: Generative AI was used to create deepfake videos falsifying the image and voice of President Maia Sandu.

Indirectly caused: The malicious content was shared across Telegram channels and sponsored on anonymous Facebook pages to reach a wide audience.

Inferred additional harm: It is likely that other malicious synthetic media assets were generated as part of this coordinated campaign but not explicitly detailed in the report.

Democracy

Reported: The report explicitly describes the erosion of democratic norms, stating the fakes aim to 'create mistrust and division in society' and 'weaken democratic institutions'.

Directly caused: Deepfakes falsely depicted the President resigning and urging citizens to vote for a specific party two days before local elections.

Indirectly caused: The campaign undermines public trust in democratic leadership, state institutions, and the integrity of the electoral process.

Inferred additional harm: Continued unmonitored deepfake campaigns could lead to systemic degradation of democratic participation and public trust in elections in Moldova.

Epistemic

Reported: The report explicitly describes epistemic harm through the fabrication of the President's image and voice to spread false statements.

Directly caused: The AI system generated highly realistic but entirely fabricated video and audio of Maia Sandu making statements she never made.

Indirectly caused: The dissemination of these fakes pollutes the information ecosystem, making it difficult for citizens to verify official communications.

Inferred additional harm: The erosion of a shared consensus reality regarding official government statements among the Moldovan population.

People affected

• Occurrences reported: 3
• People reportedly harmed: 1
• People reportedly exposed: 100000

Potential causes

Management

• Slow Platform Response: Social media companies failing to act on government requests.

Technology

• AI Deepfake Synthesis: AI tools used to falsify image and sound of the president.
• Accessibility of AI Tools: Indiscriminate availability of advanced AI generation tools.

Data Inputs

• Presidential Media Scraped: Real video and audio of the president used to train deepfakes.

Human Factors

• Malicious Actors: Fugitive oligarch and foreign actors spreading disinformation.
• Audience Susceptibility: Russian-speaking audience targeted to create societal division.

Process and Methods

• Inadequate Platform Moderation: Social media platforms failing to block paid fake posts.
• Lack of Rapid Verification: No rapid response mechanism to tag deepfakes on platforms.

Regulatory Environment

• No National AI Strategy: Moldova lacks legislation or strategy on monitoring AI use.
• Weak Platform Regulation: Lack of enforcement on global IT platforms to curb disinformation.

Information quality

• Classification confidence: High
• Reason for confidence: The report provides clear, factual details about the deepfake videos, their political context, the platforms used for dissemination, and the targeted individual. The role of AI in generating the fakes is explicitly stated, allowing for highly confident classifications.
• Ambiguities identified: The specific OpenAI model or tool used to generate the deepfakes is not named, though OpenAI is identified as the technology provider.
• Alternative interpretations: None. The incident is clearly a politically motivated deepfake campaign.

AI-generated deepfakes of Moldovan President Maia Sandu were deployed in a coordinated disinformation campaign linked to Russian-backed actors. The incident represents a substantial threat to information security, sovereignty, and societal stability by targeting democratic processes and public trust.

• Overall national security impact: Substantial
• Scope: Single nation
• Primary target: Moldova
• Alleged perpetrator: Ilan Shor and Russian-coordinated actors

Threat characteristics

• Imminence: Near-term. Represents a developing hybrid threat with recurring incidents timed around key political events.
• Autonomy: Human-controlled. The generative AI tools were directly operated by humans to create specific deceptive media, with no autonomous decision-making by the AI.
• Novelty: Evolved capability. Represents an evolved capability of AI-enabled political impersonation integrated into coordinated foreign hybrid warfare operations.

Impact by dimension

• Physical security: Negligible. No physical threats, kinetic attacks, or critical infrastructure disruptions were reported in connection with this incident.
• Information security: Substantial. Coordinated foreign-backed disinformation campaign utilizing AI deepfakes of the President to manipulate public opinion and spread false narratives.
• Sovereignty: Substantial. Deepfakes targeted the President right before local elections and major national addresses, aiming to undermine state authority and disrupt democratic processes.
• Economic security: Negligible. No significant threats to financial systems, strategic technologies, or economic infrastructure were reported.
• Societal stability: Substantial. The campaign targeted the Moldovan public at scale to sow social division, erode trust in public institutions, and manipulate democratic consensus.