This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
The Donald Trump presidential campaign released AI-generated videos featuring deepfaked voices of various public figures to mock Ron DeSantis' campaign launch. These videos, alongside other instances of AI-generated content shared by the campaign, highlight the increasing use of generative AI in political influence operations. The report notes that the specific AI models used for the campaign's clips remain unknown.
Former President Donald Trump released two AI-generated videos using deepfaked voices to mock Florida Governor Ron DeSantis. The first video, posted on platforms like Rumble and Instagram, depicted a chaotic and offensive fake Twitter Spaces event featuring deepfaked voices of Elon Musk, George Soros, Klaus Schwab, Dick Cheney, Adolf Hitler, and a generated voice of Satan. The second video showed a rocket with "Ron 2024" written beside it falling and exploding before liftoff.
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 used AI voice cloning to impersonate and mock specific individuals, including Ron DeSantis and Anderson Cooper, for political advantage.
The campaign used generative AI to create and distribute deepfakes of public figures at scale to influence political processes and public opinion during a presidential primary election.
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 was caused by the intentional decision of humans within the Trump campaign to generate and distribute deepfaked media to influence public opinion.
Added categories peak at 3: Substantial (Epistemic Harm, Inferred)
The Donald Trump presidential campaign utilized generative AI voice-cloning tools to release deepfaked videos mocking political rivals during the 2024 primary. While intended as parody, the incident highlights the evolving threat of synthetic media in domestic political campaigns, contributing to epistemic harm and the erosion of democratic norms, though posing minimal direct national security risk.
The AI systems (voice cloning and image generation models) were used as tools by the campaign to perform specific, well-defined functions based on direct human inputs, without exhibiting independent planning or autonomy.
Risk Level 3: Limited Risk. The incident involves the creation and dissemination of AI-generated content such as deepfakes, which are subject to transparency obligations under the EU AI Act to ensure users are aware the content is synthetic.
The primary AI technology used in the main incident was voice cloning to generate synthetic audio of various public figures, while image generation (Midjourney) was used in other campaign materials mentioned.
The report provides clear, detailed accounts of the specific deepfaked videos and images released by the Trump campaign, identifying the public figures targeted and the platforms used. While the exact technical models for the voice cloning are unspecified, the nature of the AI involvement (deepfakes and image generation) is unambiguous.
The content could be viewed purely as protected political parody rather than a malicious disinformation campaign, though its potential to mislead still poses epistemic risks.
Technology
Advanced Voice Cloning Models
Allowed highly realistic replication of public figures' voices without consent.
Generative Video Tools
Enabled creation of fake campaign videos like the exploding rocket.
Lack of Platform Detection
Social media lacked tools to flag or block synthetic media automatically.
Data Inputs
Abundant Public Audio Data
Abundant recordings of public figures enabled training of accurate voice models.
Unlabeled Synthetic Content
Lack of metadata or watermarks on generated assets facilitated dissemination.
Human Factors
Intentional Deception by Campaign
Campaign staff deliberately used AI to mock opponents and influence voters.
Public Susceptibility to Fakes
Audience struggles to distinguish real audio from synthetic political parody.
Process and Methods
Inadequate Content Moderation
Platforms lacked robust moderation processes to handle political deepfakes.
Absence of Verification Checks
No pre-upload verification processes for synthetic political advertisements.
Management
Prioritizing Engagement Over Truth
Campaign management prioritized viral engagement over ethical communication.
Lack of AI Ethics Policies
Campaigns lacked internal guidelines or oversight for using generative AI.
Regulatory Environment
Lack of Deepfake Regulations
Absence of clear laws banning or restricting political deepfakes in campaigns.
Weak Platform Accountability
No regulatory penalties for hosting unlabeled AI-generated political content.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
