Researchers from the University of Stuttgart conducted a human-subject study to understand how people perceive AI-generated disinformation. They presented participants with news fragments and asked them to classify their origin (human or machine) and veracity (real or fake). The results were analyzed using an adapted cybersecurity kill chain, which is typically used to describe the stages of a cyberattack. The study found that people often struggle to distinguish between human-generated and AI-generated content, even when they are suspicious. Additionally, the researchers discovered that modern language models can produce text that is indistinguishable from human-written text. This has implications for the development of effective defenses against AI-driven disinformation. --- Why it matters: These findings matter because they highlight the challenges of detecting AI-generated disinformation and the need for more proactive defense strategies. Engineers working on natural language processing and machine learning will be interested in understanding how to improve the accuracy of these models and develop more effective countermeasures. Source: https://arxiv.org/abs/2608.21389