The study by Google’s DeepMind division found that artificial intelligence-generated deepfakes impersonating politicians and celebrities are more common than AI being used for cyber attacks. The most prevalent misuse of generative AI tools is creating fake images, videos, and audio, followed by falsifying information using text-based tools. The main goal of misusing generative AI is to shape public opinion, raising concerns about the influence of deepfakes on elections. Deepfakes of UK Prime Minister Rishi Sunak and other global leaders have appeared on social media platforms, potentially impacting voter behavior. The study highlights the risk of distorting sociopolitical reality with AI-generated content and the long-term threats to democracies. As generative AI tools become more widely used, companies are monitoring the spread of misinformation and harmful content. Google DeepMind’s research aims to improve evaluations for model safety and raise awareness of the harms caused by misuse of generative AI tools. The study analyzed around 200 incidents of misuse, showing that most incidents involve easily accessible tools and minimal technical expertise, making it easier for bad actors to misuse generative AI.
Source link
Source link: https://www.ft.com/content/8d5bc867-c69d-44df-839f-d43c92785435
Top list of malicious AI use: Political deepfakes revealed. #DeepMind