Manipulated Deepfake Video of Lai Ching-te Endorsing Rivals in Lead-up to January Presidential Elections

In the lead-up to Taiwan's presidential election in January 2024, a deepfake video circulated showing candidate Lai Ching-te endorsing his rivals. Taiwanese intelligence issued warnings of intensified Chinese disinformation campaigns, such as Spamouflage, aimed at manipulating the election outcome.

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 to conduct a large-scale disinformation campaign aimed at manipulating political processes and public opinion in Taiwan.

Additional risk subdomains

• 3.1 False or misleading information: The AI system was used to generate specific false videos of political figures to mislead the public.

Causal factors

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

The deepfakes were intentionally created and distributed by human actors to influence the Taiwanese election using post-deployment AI generation tools.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk: The report describes the creation and dissemination of 'AI-generated content such as deepfakes', which falls under Limited Risk due to transparency and disclosure obligations.

AI system and alleged parties

• AI system: unspecified
• AI purpose: Deepfake Video Generation; Social Media Content Generation
• Behaviour type: Tool
• Alleged developer: People's Liberation Army, Chinese Communist Party, Base 311
• Alleged deployer: People's Liberation Army, Chinese Communist Party, Base 311
• Alleged harmed parties: Taiwanese voters, Lai Ching-te, Electoral integrity, Democratic Progressive Party, 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 Severe
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Negligible, indirect Minor
• Democracy: direct Minor, indirect Minor
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Minor
• Epistemic: direct Minor, indirect Severe
• 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 political figures.

Directly caused: Deepfake videos of Lai Ching-te and Tsai Ing-wen were generated and circulated on social media platforms.

Indirectly caused: The deepfakes were amplified by official Chinese media and shared within online chatrooms and social media networks.

Inferred additional harm: It is highly likely that numerous other unmentioned deepfakes and synthetic media pieces were generated and circulated as part of the broader election interference campaign.

Civil rights

Reported: The report does not explicitly describe violations of human or civil rights, though it touches on election interference.

Indirectly caused: The manipulation of public opinion via deepfakes indirectly threatens the civil right of citizens to participate in free and fair elections.

Democracy

Reported: The report explicitly describes efforts to undermine democracy, manipulate public opinion, and interfere in Taiwan's presidential election.

Directly caused: Deepfakes of presidential candidates were used to mislead voters and disrupt the democratic election process.

Indirectly caused: The campaigns aimed to erode public trust in democratic institutions, the government, and the integrity of the election.

Inferred additional harm: The cumulative effect of such cognitive warfare campaigns is a long-term erosion of public faith in democratic processes and institutions across the entire Taiwanese electorate.

Psychological

Reported: The report explicitly mentions that online campaigns aimed to instil fear in the public regarding potential war.

Indirectly caused: The dissemination of alarmist propaganda and deepfakes likely caused anxiety and fear among members of the Taiwanese public regarding national security.

Inferred additional harm: Widespread mild psychological distress, anxiety, and fear are likely to have been induced across the electorate due to coordinated cognitive warfare, though unquantified.

Epistemic

Reported: The report explicitly describes epistemic harm through the generation and spread of AI-generated deepfakes and false narratives.

Directly caused: AI-generated deepfakes fabricated statements from Lai Ching-te endorsing rivals and Tsai Ing-wen promoting crypto.

Indirectly caused: The spread of these deepfakes contributed to a polluted information ecosystem where citizens struggle to distinguish real campaign footage from synthetic hoaxes.

Inferred additional harm: The widespread availability of deepfake tools likely led to broader skepticism of all digital media, creating a 'liar's dividend' where real events can be easily dismissed as fake.

People affected

• Occurrences reported: 2
• People reportedly harmed: 2
• People reportedly exposed: 23500000

Potential causes

Management

• State-Sponsored Influence Operations: Foreign ministries coordinate systematic cognitive warfare campaigns.
• Insufficient Platform Moderation: Social media management fails to proactively block deepfake distribution.

Technology

• Deep Learning for Deepfakes: Deep learning models allow creation of highly realistic video and voice hoaxes.
• Voice Cloning Technology: Realistic voice synthesis mimics political figures to deliver false messages.

Data Inputs

• Authentic Video/Audio Footage: Real media of candidates is manipulated to generate deceptive deepfakes.
• Language and Cultural Indicators: Use of simplified characters and terms reveals foreign origin of data inputs.

Human Factors

• Susceptibility to Cognitive Warfare: Public vulnerability to emotional themes like war, food safety, and fear.
• Socio-Political Polarization: Existing societal divisions make people more likely to share biased deepfakes.

Process and Methods

• Algorithmic Amplification: Social media platforms rapidly amplify sensationalist deepfake content.
• Infiltration of Trusted Groups: Actors build trust in non-political groups before introducing disinformation.

Regulatory Environment

• Lack of Deepfake Regulation: Absence of strict global regulations on generating and sharing synthetic media.
• Geopolitical Gray-Zone Tactics: State actors exploit regulatory gaps in international cyber laws to interfere.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide clear, detailed accounts of the deepfakes used in the Taiwanese election, backed by statements from intelligence authorities, fact-checking organizations, and political figures. The role of AI is explicitly identified as deep learning used to create realistic hoaxes.
• Ambiguities identified: The exact origin and creators of the deepfakes are attributed to Beijing/China by Taiwanese authorities, but direct technical attribution (IP addresses) is noted as difficult to establish.
• Alternative interpretations: The deepfakes could be viewed as isolated cyber-pranks or financial scams (in the case of the crypto video) rather than a coordinated state-sponsored cognitive warfare campaign, though the evidence strongly points to the latter.

AI-generated deepfakes of key political figures were deployed in a coordinated, state-sponsored cognitive warfare campaign targeting the 2024 Taiwanese presidential election. While the immediate impact was mitigated by robust local fact-checking networks, the incident highlights the substantial and evolving threat that AI-driven information operations pose to democratic sovereignty and electoral integrity.

• Overall national security impact: Substantial
• Scope: Single nation
• Primary target: Taiwan
• Alleged perpetrator: China

Threat characteristics

• Imminence: Long-term. The incident represents an ongoing strategic and capability concern regarding foreign election interference rather than an active, immediate crisis.
• Autonomy: Human-controlled. AI systems were used as tools by human operators to generate specific synthetic media assets, with no autonomous decision-making by the AI.
• Novelty: Evolved capability. While deepfakes are an established technology, their coordinated, state-sponsored deployment at scale during a high-stakes presidential election represents an evolved threat capability.

Impact by dimension

• Physical security: Negligible. The incident involved purely digital synthetic media and disinformation with no physical threat or impact on critical infrastructure.
• Information security: Substantial. State-sponsored actors deployed highly targeted deepfakes of political leaders as part of a coordinated cognitive warfare campaign to influence public opinion.
• Sovereignty: Substantial. The AI-generated disinformation targeted the core democratic process of a presidential election, attempting to manipulate the electorate and erode trust in government.
• Economic security: Negligible. No significant strategic technology theft or widespread financial system impacts were reported from this incident.
• Societal stability: Minor. The campaign aimed to manipulate public opinion and instill anxiety, but robust fact-checking networks and civil society resilience prevented major societal destabilization.