The platform analyzes several types of content at once – images, video, audio and text. Algorithms look for characteristic signs of machine generation: from visual artifacts and noise features in images to anomalies in the dynamics of video frames, voice patterns and statistical deviations in texts. Based on this data, the system generates an estimate of the probability that the material has been created or significantly altered with the help of AI, AntiMalware writes.
The developers emphasize that the service is aimed not only at specialists, but also at a wide audience – journalists, bloggers and ordinary users who need to quickly assess the reliability of dubious content. The company notes that the service was conceived as a tool for “digital hygiene” against the background of the growing number of fraudulent schemes using dipfakes.
In particular, we are talking about cases when attackers use fake video or audio messages, imitating acquaintances in order to extort money or make people click on malicious links. Against this background, the tool is positioned as an additional filter before publishing or forwarding controversial materials.
Architech AI claims that the accuracy of detecting signs of AI generation on tests reached 98.03%. At the same time, the developers stipulate that the result of the service is not a final verdict, but serves only as an indicator that should prompt additional verification.
