Garrity campaign uses AI to impersonate Shapiro, delaying disclosure until near the end

Stacy Garrity’s campaign for Pennsylvania governor has put an AI-generated Josh Shapiro into an attack ad that waits about 45 seconds to identify its artificial origin. The delayed disclosure has drawn criticism, but Pennsylvania has no law prohibiting political deepfakes or requiring an AI disclaimer, WPSU reported on October 27.

The opponent becomes the messenger

The video opens with an apparent Shapiro smiling as he opens the doors to the governor’s office. Behind him are a desk and the American and Pennsylvania flags. The setting presents the synthetic figure as the Democratic incumbent, rather than as a narrator describing him.

The generated figure introduces himself as Josh Shapiro and begins an attack involving sexual-harassment scandals. He then points to a real New York Times article headlined “Shapiro Faces Scrutiny Over Sexual Harassment Complaint Against Aide.” That distinction is important: the cited headline concerns a complaint against an aide, not a confession by Shapiro.

There is no watermark or on-screen text identifying the video as artificially produced. Instead, the disclosure comes through the impersonation itself: near the end, the synthetic Shapiro says the ad was created with AI, adding a reference to “big beautiful data centers.” Viewers hear the attack before that explanation arrives.

A warning after the message

Anthony Moretti, a communications professor at Robert Morris University, focused on the timing in an interview with CBS Pittsburgh. He argued that limited attention spans make a late acknowledgment consequential: the campaign’s message has already been delivered by the time the video explains how it was made.

By 40 or 45 seconds, the message is pretty well-delivered: that 'I, Josh Shapiro, am a bad guy,' but obviously, that's not Josh Shapiro

Anthony Moretti, Robert Morris University communications professor, speaking to CBS Pittsburgh

Garrity, the Republican challenger, told KDKA-TV that she stands by the ad. Shapiro’s campaign took the opposite position, describing it to Axios as “cheap, dishonest, AI BS,” according to CBS Pittsburgh. The dispute is not over whether AI was used—the ad eventually acknowledges that—but over the use of the opponent’s fabricated performance.

Pennsylvania’s proposals have not become rules

Pennsylvania lawmakers have considered restrictions, but those efforts have not produced a political-deepfake prohibition or disclosure requirement. WPSU reported that a separate campaign-deepfake bill passed the state House last year but did not clear the Senate. House approval alone did not put that restriction into effect.

In May, state Sen. Lisa Boscola circulated a memo outlining another approach: prohibit AI-generated or AI-altered video, audio or images of a candidate during the 90 days before an election. As of WPSU’s October 27 report, that bill had not been introduced. It remains a stated legislative plan, not a restriction governing the Garrity ad.

Nearby Delaware, Maryland and New York have laws that either prohibit election deepfakes or require an AI disclaimer, WPSU reported. Those are different regulatory choices: one restricts the synthetic election material itself, while the other requires viewers to be told what they are seeing. Boscola’s outlined proposal takes the prohibition route during a defined pre-election period.