Two synthetic news presenters delivered political messages in videos that Graphika linked to Spamouflage, a pro-Chinese organization. The videos appeared on social platforms, including Facebook, Twitter, and YouTube, and may mark an early confirmed use of deepfake video for political propaganda, according to the report.
How the videos were put together
Graphika said Spamouflage almost certainly began distributing AI-avatar videos in late November 2022. The clips featured a male and a female presenter under the name Wolf News, a media brand the group had reportedly used before.
The presenters, called “Anna” and “Jason,” were based on recordings of real actors. Their original voices were replaced with synthetic speech in different languages, while AI altered their lip movements and facial expressions to fit the new words.
The actors also appear in other online videos made by companies using Synthesia, a UK-based provider of AI video software. Graphika said the propaganda clips were almost certainly made with Synthesia’s tools. That attribution is presented as an assessment, not a confirmed account of who created the videos.
Political messages in a news format
The videos combined archival pictures and news footage found online with synthetic narration. The format gave the messages the appearance of a short news segment, with two presenters speaking directly to viewers.
The reported topics included criticism of what the presenters described as repeated empty rhetoric about gun violence in the U.S. They also argued for U.S.-China cooperation in restoring the global economy, emphasizing China’s role.
That presentation matters because viewers encounter political claims through a familiar broadcast format, even though the presenters are AI-modified and the footage is assembled from online material. The video can make a message feel packaged as news without establishing that the claims themselves have been independently verified.
What the report says about AI and influence campaigns
Graphika described the suspected use of commercial AI tools as a way for influence operations to produce deceptive content faster and at greater scale. The report also said other actors adopted similar tactics in the weeks after Graphika identified the activity.
One example involved unidentified actors using the same AI tools to make videos aimed at online conversations in Burkina Faso. The report does not identify those actors, so the example illustrates the spread of a method rather than proving that the same organization was responsible.
Synthesia’s stated rules restrict the service to commercial use with strict moderation. Its website says, “Political, sexual, personal, criminal and discriminatory content is not tolerated or approved.” The videos described by Graphika would conflict with those rules if the reported use of Synthesia was correct.
A significant claim, with uncertainty attached
Graphika said that, if its suspicions were correct, the case would be the first confirmed use of deepfake videos for political propaganda. That qualification is central: the report makes a strong attribution, while acknowledging that it rests on analysis of the videos and their distribution.
A deepfake video of Zelenskyy from last year was also suspected of being AI-manipulated, but that could not be verified so far. The distinction between a suspected manipulation and a confirmed use is why the report’s wording remains cautious.
Graphika published its findings in a report titled “Deepfake It Till You Make It.” The videos described in the report can be viewed online at The New York Times. Taken together, the case shows how AI avatars, synthetic voices, and reused footage can be combined to give political messaging the surface of a news broadcast.