SVT Falls for AI-Generated Deepfake Animal Footage in Stress Test
Swedish public broadcaster SVT aired manipulated clips as part of a documentary experiment.
1Broadcaster tricked by AI
Several local SVT newsrooms in Sweden published deepfake videos of animals, believing them to be genuine footage. The clips featured a lynx in Ystad, a wolverine in Gävle, and an otter in Västerås. These segments were part of an unauthorized field experiment designed to test the broadcaster's internal verification procedures.
2The fake source Gustav
A person using the alias Gustav contacted multiple SVT regional offices, providing tips about rare animal sightings. The newsrooms conducted interviews with this individual, which were also deepfaked. While some editorial teams grew suspicious and declined to publish the content, three regional offices proceeded to broadcast the manipulated material publicly.
3Documentary experiment revealed
The deception was orchestrated by researcher and presenter Gustav Almqvist for an upcoming documentary series. The project aims to conduct field experiments that stress-test Swedish society's readiness against AI-related threats. Almqvist specifically targeted his own employer, SVT, to evaluate how effectively their newsrooms identify and handle sophisticated digital misinformation.
4Internal verification failure
The incident highlights significant vulnerabilities in newsroom verification workflows when faced with high-quality synthetic media. Although the clips were removed shortly after publication, the event exposed a gap between standard journalistic practices and the evolving nature of AI-driven deception. SVT management acknowledged the error, emphasizing the need for transparency.
5Justification of the method
Axel Björklund, the program's responsible publisher, defended the experiment despite the potential reputational damage to the broadcaster. He argued that testing internal routines against real-world risks is necessary to ensure institutional resilience. The team believes that exposing these weaknesses is essential for preparing the organization for future digital threats.
6SVT response and transparency
Jessica Wennberg, SVT's program area manager for Sweden, addressed the blunder by highlighting the importance of admitting mistakes. She stated that the organization must remain open and transparent whenever journalistic processes fail. The broadcaster is now reviewing how these specific clips bypassed their established verification protocols during the production process.
7Role of SVT Verifierar
SVT utilizes a specialized unit called SVT Verifierar to combat misinformation by combining traditional reporting with digital forensic analysis. This team analyzes images, metadata, and open-source intelligence to confirm the authenticity of viral content. The recent failure serves as a case study for the unit to improve their detection capabilities.
8Broader implications for media
This incident serves as a stark reminder of the challenges news organizations face in the era of generative AI. As deepfakes become more accessible, the threshold for verifying user-generated content must rise significantly. The experiment underscores that even established media institutions remain susceptible to well-crafted, AI-enabled social engineering tactics today.
Original article: SVT Nyheter — SVT Falls for AI-Generated Deepfake Animal Footage in Stress Test