Defense and Intelligence sectors. More recently, a broader range of government services, including other sensitive domains such as taxation, environmental protection, and voter verification and registration, have also become increasingly technology-dependent. 62 Governments now have unprecedented access to data on citizens – and technology companies often have even better access than the governments themselves do. 63 As the computing power available to governments and technology companies continues to rise, it becomes easier for both entities to monitor citizens’ activities. When managed responsibly, analysis and processing of citizen data enables governments and the technology companies with whom they work to enhance public services. This can remain beneficial for citizens if effective legal guardrails are in place and both governments and technology providers act in ways that earn trust. 64 However, without these conditions, the risks of misuse of surveillance capabilities rise. 65 There is divergence worldwide around how governments can use the data that they can access, reflecting ideology and culture, as well as the technological capacity and resources available to each government. Regulations, such as the European Union’s General Data Protection Regulation (GDPR) also play a role, aiming to enhance personal data protection by placing stricter limits on data usage by governments and businesses. Meanwhile, citizens often remain unaware of how their personal data is collected, used and shared, limiting their ability to make informed decisions. Figure 1.24 shows the close connectivity between Censorship and surveillance, Societal polarization, Misinformation and disinformation and Online harms, highlighting the confluence of these risks in the digital ecosystem.Censorship and surveillance ranks #16 in the GRPS risk ranking on a two-year outlook, increasing five positions since last year, showing that concern respondents have around this issue is real and growing. In a world of deepening societal and political divisions, amplified by eroding trust in the digital environment, concerns with Censorship and surveillance are most pronounced in Eastern Asia, Latin America and the Caribbean, and Central Asia, according to the EOS (Figure 1.25). Notably, Nicaragua ranks this risk as the fourth-most severe threat over the next two years, while eight other economies identify it among their top 15 risks. Actions for today A. Expand upskilling for people building andusing automated algorithms Organizations should use AI models that minimize bias and mitigate unintended consequences in content creation and distribution. While technical solutions for significantly debiasing automated algorithms already exist, their consistent application remains a challenge. If implemented correctly, these solutions could greatly reduce the risks associated with model bias. Common debiasing strategies include data pre-processing before training a model, in-processing techniques during training, and post-processing steps after training. 66 These methods help ensure that AI models are fairer and more equitable. However, due to the rapid pace of change in AI development and the increasing complexity of its applications, keeping up with the latest advancements in algorithmic debiasing is difficult for many involved in building and using automated algorithms. To address this, there is a pressing need for continuous upskilling of developers, data Victoria Prymak, Unsplash Global Risks Report 2025 38