In April 2017, Deputy Defense Secretary Bob Work established an Algorithmic Warfare Cross-Functional Team that launched Project Maven, an initiative designed to integrate artificial intelligence (AI) and machine learning into military intelligence and combat operations. 2017 was a crucial inflection point for AI technologies, defined by a transitional period from early development stages towards large-scale integration. Project Maven served as a pioneer for AI solutions in the warfare landscape. It now sits among the preceding 800 AI projects implemented by the Pentagon, 300 machine learning tools developed by the CIA, and numerous initiatives by international organizations spanning from bodies of the United Nations (UN) to various non-governmental organizations (NGOs).
Due to its growing spotlight on the international stage and groundbreaking use cases, AI technologies have been adopted by organizations such as the World Food Programme (WFP), International Rescue Committee (IRC), and United Nations High Commissioner for Refugees (UNHCR), which represent the growing adoption of such technologies in the humanitarian landscape. Humanitarian organizations have been particularly drawn to the efficiency benefit that AI solutions provide in crisis management contexts; however, a major concern lies in the looming uncertainty surrounding its ability to keep up with the ever-changing nature of humanitarian ethical standards.
On October 9, 2023, the Israeli military completed a “complete siege” on the Gaza Strip, leading to extensive destruction and displacement. From cutting off civilian access to food, electricity, and fuel supplies to the infrastructure collapse of local hospitals, communication systems, and transportation means, the humanitarian impacts of this operation were catastrophic. The conflict resulted in the killing of 9,000 people primarily by means of airstrikes, hunger, or disease, left 25,000 injured, and 70% of residents displaced. These statistics underscored the need for improved disaster mapping, which is where AI capabilities have presented an unprecedented edge in accurately attending to crisis zones.
One of the most renowned and heavily utilized capabilities of AI is its precise mapping of disaster sites. This precision is made possible by machine-learning AI models that process satellite images taken before and after an event, programmed to highlight geospatial differences between the two. This analysis is then retrieved by humanitarian organizations such as the United Nations Satellite Centre (UNOSAT) and used to identify areas in need of reconstruction and inform actionable strategies. In Gaza, these types of reports were able to track the precise locations of destroyed infrastructure and map out civilian displacement patterns, proving to be an essential tool for accurately delegating aid directives and overall improving the efficiency of the damage mitigation process.
However, the advantages of AI satellite image tracking do not come without parallel ethical concerns, mainly regarding the violation of civilian privacy rights. In recent years, the United Nations General Assembly (UNGA) has expressed concerns surrounding the rapid pace of AI technological development, with beliefs that it will inherently enhance the capacity of governments, companies, and individuals to undertake surveillance, interception, and data collection. The UNGA characterized “unlawful or arbitrary surveillance” as any “highly intrusive act” which violates privacy rights in non-consensual situations.
In cases of AI satellite imagery, current high-resolution technology satellites can identify features as small as 31 centimeters (about one foot), implying its ability to monitor the precise movement of individuals or groups, recognize faces, and generate detailed images of private property. However, an important line must be drawn between the credible usage and perceived exploitation of such sensitive data points. Adopting AI technology presents a nuanced risk for humanitarian organizations, who must carefully balance advantageous opportunity with moral caution; failure to consider ethical standards in practice may ironically threaten civilian welfare.
In addition to post-disaster relief, AI technology has been proven useful in providing pre-disaster aid to endangered communities using machine learning models in tandem with cloud-based data processing tools. For example, AI companies have designed software such as Google’s Flood Forecasting System, which analyzes weather patterns, and California’s Earthquake Warning System, which monitors seismic activity, to predict future occurrences of natural disasters. Such technology’s unprecedented accuracy in predicting notoriously erratic events has proven to be an invaluable asset for humanitarian organizations’ proactive resource allocation initiatives.
Over the past few years, the UNHCR has been utilizing AI to build forecasting models that anticipate refugee movements, inform planning, and provide guidance for resource allocation. Their 2022 model, Project Jetson, which was built on climate, remittance, and market price data sources, functioned to predict levels of forced displacement in Somalia and preemptively respond to anticipated violence and conflict escalation accordingly. Similarly, the WFP has developed a model that projects food insecurity levels in international conflict zones with a mission of understanding the trajectory of and responding to anticipated cases of undernourishment.
