National risk perceptions: Misinformation and disinformation FIGURE 1.23 Source World Economic Forum Executive Opinion Survey 2024.Executive Survey Opinion rank of national risks from the question “Which ﬁve risks are the most likely to pose the biggest threat to your country in the next two years?” Rank 1st10th20th31st images, voice or text – can be difficult. GenAI lowers the barriers for content production and distribution, and some of that content is inaccurate. Threat actors, state agencies in some countries,49 activist groups, and individuals who may or may not have criminal intentions can automate and expand disinformation campaigns, greatly increasing their reach and impact.50 Misinformation and disinformation can also be the result of AI-hallucinated content or human error, and these too are likely to rise amid the growing volume of content. The upshot is that it is becoming increasingly hard to know where to turn for true information. Both political and Societal polarization skew narratives and distort facts, contributing to low and declining trust in media.51 Across a sample of 47 countries, only 40% of respondents said that they trusted most news.52 According to the EOS, respondents in high-income countries are generally more likely to express concern about the risk of Misinformation and disinformation over the next two years than respondents in lower-income countries, with some exceptions. This risk ranks among the top five in 13 countries, including India, Germany and Canada, and features in the top 10 in 30 additional countries (Figure 1.23). Respondents identifying this risk often also highlight Societal polarization as one of the most severe risks in the same timeframe. Poor quality content and lack of trust in information sources continue to present a threat to societies.53 Algorithms, especially complex machine learning models, can also be an entry point for cyberattacks that use disinformation. An example of this would be a structured query language injection attack, in which inputs are manipulated to generate incorrect outcomes or to compromise training data sets.54 As many models lack transparency, either by intention, by accident, or because of intrinsic opacity, it is difficult to identify vulnerabilities and mitigate potential threats. In addition, given the reliance of algorithms on third-party data sources, software libraries and network infrastructures, threat actors can compromise the supply chain to manipulate algorithms and cause widespread damage. Further, as algorithms come to govern or influence more aspects of society, the potential for coordinated cyberattacks using automated systems grows. Global Risks Report 2025 35