This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
An AI-generated political campaign ad in Pennsylvania used a deepfake likeness of Governor Shapiro to deliver a misleading message about a sexual harassment complaint, raising concerns about election manipulation.
Pennsylvania gubernatorial candidate Stacy Garrity's campaign reportedly released an AI-generated video impersonating Gov. Josh Shapiro and placing first-person statements about harassment scandals, intimidated accusers, and hush money in his mouth. The harassment allegations concerned a former aide. Shapiro's campaign denounced the ad as dishonest; a media scholar warned that its delayed AI disclosure could mislead viewers.
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 ad generated discrete misleading content by using a synthetic likeness to falsely associate the governor with a sexual harassment complaint.
The AI system was used to generate a deepfake of a political candidate to influence public opinion and manipulate political processes during an election campaign.
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 risk was caused by the intentional decision of humans (the political campaign) to create and deploy a deepfake ad using a deployed AI tool.
Added categories peak at 2: Minor (Epistemic Harm, Direct)
An AI-generated political campaign ad used a deepfake likeness of Pennsylvania Governor Shapiro to deliver a misleading message shortly before Election Day. This incident represents a substantial threat to information security and democratic processes, reflecting an evolving trend of domestic political actors leveraging synthetic media to manipulate voters.
The AI system was used as a tool by the campaign to generate a synthetic video and audio likeness of the governor based on direct inputs.
Limited Risk: The report describes an AI-generated campaign ad using a deepfake likeness, which falls under the category of AI-generated content such as deepfakes that require transparency disclosures.
The AI system was used to generate a realistic synthetic likeness (deepfake video and voice) of Governor Shapiro for a political advertisement.
The report clearly details the creation and deployment of the AI-generated ad, the reactions of both campaigns, and the broader context of deepfakes in elections. There is little ambiguity about what occurred.
None
Technology
Realistic Deepfake Generation
AI tools easily generate convincing likenesses of political figures.
Delayed Disclaimer Timing
The ad hides the AI disclosure until 45 seconds into the video.
Data Inputs
Unrestricted Training Data
Publicly available media of politicians is used to train AI models.
Human Factors
Short Audience Attention Spans
Viewers absorb the misleading message before seeing the disclaimer.
Campaign Intent to Mislead
Political actors willingly use deceptive AI to attack opponents.
Process and Methods
No Mandatory Ad Labeling
No standardized process forces immediate labeling of AI content.
Management
Prioritizing Win Over Ethics
Campaign managers chose to use misleading AI to gain an advantage.
Regulatory Environment
Absence of Federal Laws
No federal legislation restricts the use of deepfakes in elections.
State Regulatory Gaps
Pennsylvania lacks active laws prohibiting deepfake political ads.
AI system involved
An AI system was involved, but the reports do not identify it.
Identified by automated extraction; can differ from the alleged developers above.
