Elon Musk posted a video on X (formerly Twitter) in which an AI-generated voice imitates Democratic U.S. presidential candidate Kamala Harris. Musk did not indicate that the video was fake or a parody.
The video uses AI to mimic Harris' voice, making the vice president say things she did not say in the original video. The deceptively real fake Harris voice claims that President Biden is senile, that she herself has no idea how to run the country, and that she is the "ultimate diversity hire" as a woman and person of color.
In addition to the audio, the visuals were also altered: Images of former President Donald Trump and his running mate JD Vance were removed and replaced with images of President Biden.
The video's creator, X user "@MrReaganUSA," labeled it an "ad parody" in his original post. But Musk shared the video without the parody disclaimer, saying "This is amazing," followed by a laughing emoji. The video has well over 100 Million views so far.
The video appears to violate X's guidelines, which state that synthetic media that has been "significantly and deceptively altered, manipulated, or fabricated" must be flagged or removed.
This rule applies to videos that are "likely to result in widespread confusion on public issues, impact public safety, or cause serious harm," and is meant to be enforced especially if the videos were manipulated through artificial intelligence algorithms. "@MrReaganUSA" confirmed on X that he had used AI to clone Kamala Harris' voice.
AI manipulation of media, particularly to sway public opinion, is often cited as a major risk of current AI technologies. It undermines trust in audio and visual content.
X has not yet responded to the video. Musk's post has not been visibly corrected through "Community Notes," X's user-driven fact-checking system, and Musk himself has not publicly addressed the criticism of sharing the deepfake.
Following a report by the New York Times, "@MrReaganUSA" posted that he had clearly marked the post as a parody. He believes Elon Musk didn't even think it could be mistaken for a real campaign ad because of the obvious satire.
AI audio deepfakes have advanced rapidly over the past year. Many commercial and open-source tools can now create realistic voice clones that repeat given text in a specific person's voice.
While voice cloning has some positive uses, such as helping those who've lost their voices, its potential for abuse seems far greater. The Harris example is just more evidence of this.
