Generative artificial intelligence (GenAI) comes with various types of risks. Chenchen Wang from University of Maryland studied articles published about it in Australian digital news through a mixed-methods content analysis of 379 articles from seven outlets. The time period for the articles was from 2020 to March 2024.
The rapid development of GenAI has drawn attention to potential risks, including scholarly attention (Sætra and Danaher, 2025; Seghid et al., 2026). Here, GenAI is defined as AI with content-generation functions, such as text, images, audio, video, or code. Media framing of AI typically emphasizes progress-oriented and economic frames.
In this study, the types of risks are looked at: which risks are the most prominent, the differences in coverage between private and public outlets, the originality and geographic context of reporting, and qualitative examples of presentation of the technology. The four research questions were:
RQ1: In Australian digital news coverage, which types of GenAI risks are prioritized, and which remain relatively peripheral?
RQ2: How does the representation of GenAI risks differ between public and private media in the sample?
RQ3: How do reporting originality and geographic context vary across public and private media in GenAI risk coverage?
RQ4: How do Australian digital news articles identify and explain GenAI technologies, and how do narrative choices shape the presentation of associated risks?
The seven outlets chosen were: News.com.au, 7news.com.au, Nine.com.au, Daily Mail Australia, and The Sydney Morning Herald which were the five private outlets, and ABC news and SBS News, which were public. They were coded inductively.
Ethical, legal, and regulatory compliance risks were the most frequently reported category at 33,5%, followed by economic and job market risks at 23,7%, and then technical performance and accuracy risks at 21,9%. Policy and governance risks were at 19,5%, with education and knowledge dissemination risks at 17.4%.
Other risks that were mentioned were data and privacy security risks at 15.3%, communication and information risks at 12.4%, security and defense risks at 11.6%, and sociocultural risks at 10.8%. The least mentioned category was mental health and user behavior risks, 2,4%.
When it comes to differences between public and private media, public media was massively more focused on “policy and governance” and “education and knowledge dissemination”. “Extinction” and “technical performance and accuracy” on the other hand were more in private media. Private media was also more frequently negative tone.
Public media did far more original reporting and focused more on the Australian context. In international news, the Global South and developing countries appeared only rarely – mostly it was about the West. Mostly, the articles about GPT did not go into technical details, instead opting for terms like “AI,” “GenAI,” “GPT,” “ChatGPT,” and “chatbots”.
The distribution of risks did not follow a neat divide between long-term and short-term risks. The findings also differed from the findings in other studies focused on different countries, like Germany, where fake content was a major concern (Meissner 2024). As mentioned, technology was named more often than it was explained.
In conclusion, the results extend the scholarship to analysis on not just which risks appear, but also risks become recognizable, how the technological basis is described, and the context of the definitions.
The article “What generative artificial intelligence (GenAI) risks make the news? representation of GenAI risk in Australian digital news” by Chenchen Wang is in Media International Australia. (Free abstract).
Picture: Business and finance concept. Word RISK on blue cubes with magnifying glass by Sasun Bughdaryan.
License Unsplash.
