The mediated amplification of societal risk and risk governance of artificial intelligence: Technological risk frames on YouTube and their impact before and after ChatGPT

Over the past decade, many experts from academia, business, and civil society have warned of the risks of artificial intelligence (AI), such as large-scale job displacement, data bias against social groups, the use of AI to undermine democracy, the misuse of AI for mass surveillance and/or policing, the unregulated use of autonomous weapons, and others. However, research suggests that news media and social media representations emphasize the benefits over the risks of AI and rarely address issues of risk governance, while social media studies are still rare and largely based on Twitter/X data. This study proposes a theoretical framework of mediated risk communication based on the social amplification of risk framework (SARF), the extended parallel process model (EPPM), and framing. To understand prevalent AI-related risk frames and their role in mediated technological risk amplification/attenuation before and after the release of ChatGPT, a sample of the 501 most viewed YouTube videos about AI published between 2020 and 2023 were analyzed. Four frames with different emphases on societal threat and societal efficacy of AI were found with significant differences between risk/benefit emphases, AI evaluations, risk control/governance measures, and reference to risk governance actors. Frames with higher emphasis on societal threat of AI were more likely to be viewed and commented on by YouTube users. AI-related threat frames increased significantly after the release of ChatGPT in 2022, suggesting its role as a risk-amplifying event. Implications for theory, future research, and AI risk governance are discussed.