In almost every election, we see some form of disinformation being spread about specific politicians in order to sway votes. It is a common practice occurring every year. However, this year, with the increased accessibility of generative AI tools, it is easier for people to create deepfaked content. As the Dutch elections were held on the 22nd of November, we saw an increasing number of deepfakes of politicians being shared on the internet to mislead public opinion, potentially influencing voting behaviour.
A recent example of deepfake manipulation occurred during the Dutch elections held on November 22nd. Deepfake images of Frans Timmermans, the party leader of GroenLinks-PvdA, were circulated online, depicting him in a private airplane. Our deepfake detection tool identified the image with a 99.9% probability of being a deepfake. These fabricated images were designed to trigger negative sentiment towards Timmermans, potentially influencing voter perceptions.
However, it does not end with just Timmermans. A music video featuring several prominent political figures was widely shared online. This deepfake video altered the words and actions of the candidates. Our tool was able to detect the deepfake with an 84% probability. As we navigate the digital age, information plays a major role in shaping our choices. The ability to distinguish real from fake is important, it is important that we equip ourselves with the knowledge and tools necessary to critically evaluate the information we encounter online. This includes understanding the nature of deepfakes, recognizing their potential for manipulation, and developing strategies to identify them. By critically evaluating the sources and content we encounter online, we can empower ourselves to make informed decisions.
We created a blog post a while back on everything you need to know about deepfakes and how to spot them. You can read more about it here.
You asked, and we answered!
The emergence of Generative AI tools, like Midjourney and Dall-E, introduced a new dimension of creativity and complexity. While these tools offer immense potential for content generation, they also raise concerns about authenticity and the potential for misuse. We recognize the importance of staying ahead of these advancements and have been working tirelessly to develop a detection method to identify Generative AI-generated content, and we’re excited to announce that it’s finally here!