Artificial intelligence (AI) today occupies a growing place in the fields of diplomacy, international security and, consequently, the prevention of armed conflicts. Thanks to technological advances in machine learning (machine learning) and the analysis of massive data, certain systems are now capable of anticipating risks and recognizing patterns in order to formulate forecasts on the evolution of international conflicts.
In this context the notion of “predictive security” appears. This term refers to the use of artificial intelligence and data analysis systems, aimed at predicting and anticipating conflicts as well as the violence and instability that accompany them, so as to inform the strategic and security decisions of different actors. However, while these technologies offer promising prospects for improving conflict prevention, they also face significant structural limits that restrict their application and effectiveness.
La sécurité prédictive
The rise of artificial intelligence in the field of conflict prevention can be considered as a logical and natural evolution of recent technological progress, a response to the contemporary challenges represented by armed conflicts and their violence. However, the integration of these technologies into early warning mechanisms is not limited to the addition of a new analysis tool: on the contrary, it has the potential to gradually transform traditional modes of crisis anticipation by giving rise to an approach to security based more on algorithmic forecasting and the use of vast quantities of data. Governments and international organizations could thus better identify certain risks and preserve their resources while planning their responses.
The VIEWS model and AI for conflict prediction
« VIEWS » (Violence and Impacts Early Warning System) is an armed conflict prediction model created by the Peace Research Institute of Oslo (PRIO) and Uppsala University. It uses artificial intelligence to anticipate the likelihood of conflict escalation and the potential number of casualties around the world. One of its particularities is the free access to its forecasts.
Their motto, “stay informed, act, build resilience,” reflects VIEWS’ goal: to provide actors such as governments, international organizations and humanitarian organizations with open-access tools to quickly detect warning signs of conflict and victims of violence.
This system relies on the analysis of a large amount of data by an AI algorithm that recognizes patterns and calculates the probability of future conflicts. It is also important to note that this technology currently only analyzes one type of conflict: armed conflicts between states, although the VIEWS team plans to extend the model “to non-state conflicts and unilateral acts of violence.” These forecasts are updated every month and cover a period of 3 years. Finally, this data is translated into interactive, colorful maps or can be downloaded in its raw form.
In the VIEWS 2026 report, several countries are classified as “at risk.” Regarding the expected number of casualties and “predicted likelihood of armed conflict at the state level”, Ukraine, Sudan, Nigeria, Israel/Palestine and Pakistan had the highest scores, and this trend continues until March 2029. For example, while we know that Sudan is currently in the grip of a devastating civil war since 2023 (40,000 people killed according to the UN), VIEWS predicts that in February 2028 again, the number of deaths expected in Sudan will be 242 (864 in May 2026), and that, for this same month, the The probability that armed conflict persists will be 100% (100% in May 2026). The report thus illustrates the potential of models like VIEWS to anticipate conflict dynamics, but they also raise questions about their reliability, their limits and their implications, particularly since these ambitious predictions over several years can themselves have important human and political consequences.
The promises of predictive security
The main advantage of a system like VIEWS is its supposed ability to quickly detect changes in the dynamics of armed conflict, which could help reduce the overall number of casualties by serving as an early warning system. As the 2026 report shows us, in just one month (December 2025-January 2026), Niger went from 94% probability of armed conflict to 99%, while Tanzania fell from 85% to 64% in the same period of time. These rapid variations demonstrate the usefulness of a system capable of monitoring and predicting the evolution of a crisis almost in real time.
This capacity for anticipation could also allow international actors to better prioritize their humanitarian, military and diplomatic resources. For example, in 2025-2026 the budget for UN peacekeeping operations amounted to $5.6 billion. In this context, “human-machine collaboration” with tools like views could contribute to more effective planning of interventions by identifying the most vulnerable countries or regions and allowing more targeted use of available resources. For countries like Canada, which regularly participate in NATO missions and other international conflict prevention operations, this type of tool could prove particularly strategic.
The limits and risks of algorithmic forecasting
Despite their strategic advantages, predictive security systems like VIEWS also have several important limitations. First, armed conflicts remain profoundly human and unpredictable phenomena. Artificial intelligence models cannot therefore produce certainties but simply probable scenarios, likely to evolve very quickly depending on the context. The case of Ukraine perfectly illustrates this dynamic. In January 2026, VIEWS predicted that the number of victims in Ukraine in March 2026 would be around 3270, while the map of recorded deaths indicates 7769. This considerable discrepancy shows that even a sophisticated and award-winning model like VIEWS cannot cannot fully anticipate such sudden and unexpected events as death.
Several researchers also raise other limits of AI-based predictive security. For example, according to some recent studies, diplomacy based more on data and AI can sometimes reinforce existing biases, reproduce stereotypes or simply miss essential aspects of the causes of the conflict. These limits depend on the type and quality of data the tool is trained with and uses. Additionally, the researchers point out that the lack of transparency of some AI models can exacerbate these limitations and pose an accountability issue.
Finally, the growing use of artificial intelligence in international security also raises political and ethical issues. By identifying certain countries such as Sudan, Ethiopia, Nigeria and Pakistan as being at risk of conflict escalation over the next 3 years, a system like VIEWS can influence how the international community perceives a country, particularly those associated with this theoretical notion of “failed states”. “. This may harm the country’s prospects and reduce some opportunities, notably international investments. Indeed, the Office of the High Commissioner for Human Rights points out that the use of AI can have disproportionate impacts on groups already affected by inequalities.
Furthermore, this type of long-term prediction risks reinforcing certain dynamics of “self-fulfilling prophecies”. By persistently associating already weakened countries with a high level of risk, these predictions could influence the decisions of political, economic and international actors, therefore helping to reproduce the trajectories that they precisely seek to anticipate and avoid.
Furthermore, AI can also be used for much more controversial purposes such as mass surveillance, information warfare and the manipulation of public opinion. The risks of abuse or scope creep » therefore really exist and even a tool as open and benevolent as VIEWS could theoretically be exploited by certain ill-intentioned actors.
Towards responsible use
It is important to remember that AI and these technologies cannot truly understand the intricacies of human conflicts. Systems like the “Violence and Impacts Early Warning System†(VIEWS) produce statistical predictions based on the data provided to them, useful and rapid predictions, but which must remain a decision support tool, supervised by human experts aware of the limits of these devices, some of which have been presented in this article. This question is particularly relevant for NATO members, including Canada, in the context of the new NATO Strategy on Artificial Intelligence (2024), which aims to promote responsible and balanced use of AI in the service of the defense and security of allies.
Thus, the developments of predictive systems call for the creation of a clear and robust international ethical framework, particularly in terms of transparency, accountability and human supervision, to ensure prudent, legitimate and responsible use of these technologies in an area as sensitive as the prediction of armed conflicts.
Disclaimer:Â Any views or opinions expressed in articles are solely those of the authors and do not necessarily represent the views of the NATO Association of Canada.
Photo: Wikimedia Commons (Public Domain).
/2025/09/22/netnyahou-68d1834f9e746904592703.jpg "War in the Middle East: in Israel, the agreement between Iran and the United States is strongly criticized and weakens Benjamin Netanyahu’s strategy")
