Middle East: Using AI to stop dissent before it even starts

DW News · Feb 24, 2026 · Collected from RSS

Full Article

Imagine if back in December 2010, the Tunisian government — then headed by authoritarian leader Ben Ali — had known what would happen after a humble fruit seller set himself on fire in a public square. The gruesome death of fruit seller Mohamed Bouazizi eventually led to mass demonstrations and Ali fleeing the country. The Tunisian revolution also inspired further protests that would change the Middle East forever during the period that came to be known as the Arab Spring. But what if Tunisia's authoritarian regime had had artificial-intelligence-powered tools that predicted the coming revolution? They could perhaps have prevented other protests, or arrested leaders of a fledgling anti-government movement before it grew any larger. The Ben Ali regime didn't have those tools. But some of today's authoritarian governments in the Middle East do — at least, to a certain extent. And the potential for them to be used to suppress political dissent before it even starts is growing, as AI-powered conflict forecasting evolves. A lot of the technology used in the MIddle East comes from China, a country that already uses surveillance and AI-powered monitoringImage: CFOTO/IMAGO AI changes conflict forecasting Analysts have always tried to predict future risks. But over the past few years the availability of massive amounts of data and computing power as well as advances in machine learning mean that computing models are increasingly able to use data about past events to predict future ones. In a geo-political sense, the positives of conflict forecasting include ensuring that resources are better allocated and countries and humanitarian organizations better prepared. But researchers in the conflict forecasting community are also concerned about possible misuse of this kind of technology, confirms Christopher Rauh, a professor of economics and data science at the UK's Cambridge University and one of the founders of an organization called ConflictForecast. It uses a combination of machine learning and human expertise to "detect new, subtle signals of conflict risk in countries not currently at war." A screenshot of the ConflictForecast.org website, which works with the German and British foreign offices, among othersImage: Conflictforecast.org "Yes, we do worry that bad actors could use some of the data — that's why we don't publish predictions on protests for example," Rauh continued. "Our model is adaptable and with a little tweak, we could produce them [protest predictions] easily. But since protests can be a healthy expression of democracy, we don't publish that data — a bad actor could potentially crack down." Rauh says that at the moment conflict forecasting isn't quite fine-grained enough to be easily abused by authoritarian states. "These models work on averages when they flag high-risk situations. They are not magical," he explains. "But as prediction improves and data availability increases, there could be greater risks." How does conflict forecasting work? As with other machine-learning models — often described as artificial intelligence — the model has to first "learn" from a lot of different data. This could be anything from reports in the media to information provided by real-time conflict trackers like US nonprofit, Armed Conflict Location and Event Data or Uppsala University's Conflict Data Program, based in Sweden. It can also include economic indicators, demographics, infant mortality rates, even cell phone location monitoring and aviation movement, among any other information researchers deem relevant. One of the most important things is that there needs to be a lot of historical data, preferably going back well over a decade, Rauh explains, so that the model can detect patterns. "There is currently no AI-powered system that can perfectly predict geopolitical flashpoints or forecast their implications," researchers at UK's Alan Turing Institute, which focuses on data science and AI, wrote in March 2025. In the past, the ConflictForecast model has foreseen potential issues in Russia and Venezuela — it anticipated the Wagner group rebellion against the Russian government in 2023 and, more recently, political upheaval in VenezuelaImage: ROMAN ROMOKHOV/AFP/Getty Images The challenges include lack of computing power, data scarcity or inconsistency, and the fact that it's very hard to predict what unexpected things an individual might do — say, the Tunisian fruit seller Mohamed Bouazizi or the leaders who decided to resign rather than fight. But they may get better at that. Conflict forecasting, risk and threat analysis using AI is growing, both in the public and private sectors. This in itself could be problematic, says Damini Satija, director of Amnesty Tech, part of Amnesty International. "To work as marketed, the tools must be trained on existing data sets which means that they will recreate mistakes, biases or errors in real world data," she points out. "More critically, these tools depend on an assumption that complex and variable human behaviour can be distilled down to simple indicators. In the crime prediction or recidivism context, this has already been debunked."Why Middle East authoritarians could be first When, and if, the technology does evolve, some of the Middle East's autocracies could potentially be among countries first to use AI-powered conflict forecasting to suppress dissent. "The Middle East has been at the intersection of technological adoption with political power for a long time," Manchester University researcher Arash Beidollahkhani pointed out in a November paper for academic journal "Democratization." "Traditionally, the region's authoritarian governments have relied on surveillance, censorship and coercion ... the technologies of AI, from facial recognition to predictive analytics, have exponentially enhanced these capacities." Countries like Saudi Arabia, the United Arab Emirates, Iran, Egypt and Bahrain have already used advanced computing against opposition movements. Egypt has monitored digital communications and prosecuted activists for social media posts. The country's New Administrative Capital is being developed as a "smart city" with over 6,000 cameras on its streets, something digital rights experts have already criticized as ripe for exploitation under the current government. Saudi Arabia uses facial recognition technology for crowd management in pilgrimage hotspots Mecca and Medina, and has plans to include surveillance and emotion-recognition systems in smart cities like Neom, which is still under development. It is probably the UAE that is best positioned to use AI-powered forecasting to potentially suppress dissent. Today it is one of the most advanced in the world when it comes to using what is known as "predictive policing." Predictive policing analyzes past data to prevent future crimes, using statistical predictions to identify either locations where crimes are likely to be committed or people who are likely to commit them. The UAE is thought to be one of the most advanced countries in the world when it comes to 'predictive policing'; the others are the US (pictured), China and Japan Image: Erin Hooley/ZUMA/IMAGO The UAE already has a number of "safe city" projects running that involve the analysis of vast amounts of surveillance data, including both facial recognition and behavioral analysis. It also has the vast financial resources and private and political connections to integrate more AI-powered forecasting into Emirati society. Like other autocratic governments in the region, it doesn't need to answer to the public or be transparent about how collected data is being used. And a lot of the UAE's tools around this come from China, which is already using AI-powered technologies to suppress dissent at home. "If a government of any kind has implicit or explicit goals to crack down on dissent, there is no reason they would not attempt to use AI tools to enable this," Amnesty Tech's director Satija says. But she and other digital rights activists say that today's systems – things like predictive policing – already present a serious threat. They are already used against dissenters, Satija says, and they have "an absolutely chilling effect." For example, activists believe they will be more easily identified, so they don't ever protest. "Our greatest concerns are about existing infrastructure and what it can already enable," Satija concludes. Edited by: Andreas Illmer