In Central Africa, a small freshwater lake bordering Niger, Cameroon, Nigeria and Chad has shrunk by 90% over the past few decades. The change in water levels in Lake Chad are contributing to the rise of a new type of migrant: climate refugees. For people living on the banks of the lake, the falling water levels are an indicator of hunger, impending conflict and displacement. For some researchers, meanwhile, examining water levels, the stress the falls place on rural livelihoods and the relationship they have with conflict can indicate when large-scale forced migration might occur.

Climate change has pushed people to move from their homes and will continue to do so. Not all will be officially classed as migrants because they will not cross a national border: according to a report from the Internal Displacement Monitoring Centre, 23.7m people moved within their countries because of climate-related disasters in 2021.

What is more, climate change will exacerbate other problems that already drive people to move. For example, a White House report has found that the regions most vulnerable to climate change usually also suffer conflict and violence.

In response to these emerging challenges, some have proposed advanced digital technologies such as artificial intelligence to help governments and others to predict and prepare for migrants pushed to move by climate change.

For example, the technology entrepreneur Babusi Nyoni said in 2017 that he and his team were working on an algorithm to “intelligently suggest remedies to an impending crisis, such as that, in the event that it predicts imminent famine based on historic data and the climate change trajectory, it will recommend alternate farming methods or alternate food sourcing that will cater to a changing climate”. Nyoni said they were building an “easy to use tool for governments”.

Researchers from the Immigration Policy Lab branches at Stanford University and the Swiss Federal Institute of Technology (ETH Zurich) reported in 2018 that they had developed an algorithm to help countries place refugees in areas where they would be most likely to find employment and integrate into the host society. They said that the machine-learning tool could improve the employment prospects of refugees in the US by 41% and in Switzerland by 73%.

In 2019, the Danish Refugee Council worked with IBM Research to develop Foresight, machine-learning software that analyses historical data about politics, economics, crises and climate to predict forced displacement, with the intention of improving humanitarian planning. The council says that, to date, the tool is being used in work on some countries. Furthermore, the United Nations Central Emergency Response Fund has also used it to support funding allocation to Afghanistan and United Nations Office for the Coordination of Humanitarian Affairs in Central America.

The Danish Refugee Council says that at first the team was looking at predicting bilateral flows – how many individuals move from one country to another. But some of the problems with such an approach – for instance, it could be used to block migration rather than to facilitate it – led it to revise the model. As a result, it predicts only the number of internally displaced persons, refugees and asylum seekers, limiting the scope to movements of people within borders.

Even Facebook, Palantir and Microsoft are partnering with humanitarian agencies to support climate migrants. But privacy advocates have pointed out that turning to for-profit technology companies makes surveillance likely to proliferate yet further while failing to address the problems that force people to leave their homes in the first place.

While the potential of these tools appears promising, I am concerned about how they will be used to manage migrants. Early studies suggest that governments tend to view those in poor countries who are more likely to cross international borders and become climate migrants as undesirable. As such, the proposition that advanced digital technologies be used to predict migration crises sparked by climate change before they occur is a dangerous one: it could encourage states to put up barriers to migrants.

The reality is that migrant-receiving countries – often the same countries that largely caused climate change – also use advanced digital technologies to control migration and prevent the arrival of ‘undesirable’ migrants at their borders.

For example, the Austrian Institute of Technology’s Foldout project, funded by the European Union, uses aerial and satellite data to detect people at a border in real time. Sensors in a smart camera paired with artificial intelligence can spot people even through dense foliage.

Similarly, Croatia’s Ministry of the Interior has deployed drones for border surveillance and arrest. They can spot people almost 10 kilometres away in the daytime and 3 kilometres away at night.

In Greece, Centaur, a partly automated system, conducts surveillance of asylum camps in Lesbos, Chios, Samos, Leros and Kos to monitor crowds with cameras, drones and artificially intelligent movement analysis.

Taken together, these different initiatives suggest that European officials are more concerned with the mass control of ‘undesirable’ migrants than with protecting them.

What further complicates the question is that there is no clear and broadly accepted international agreement or definition of who is a climate migrant, let alone what rights and protections they should have. Although widely used, the term ‘climate refugee’ is misleading: according to the United Nations 1951 Refugee Convention, refugees are “people who have fled war, violence, conflict or persecution and have crossed an international border to find safety in another country”. The convention has not been updated for the era of climate change.

We already have cautionary tales from existing examples of the use of advanced digital technologies like artificial intelligence and big data technologies to respond to refugee flows.

For example, in 2020, the UK Home Office deported roughly 7,000 international students after a faulty algorithm concluded they had cheated in the English language tests they needed to pass to secure their visas.

In the US, the Immigration and Customs Enforcement agency (ICE) used an algorithm to determine if individuals arrested for immigration ‘offences’ should be detained, released on bond or trusted to appear in court. The New York Civil Liberties Union, together with The Bronx Defenders, filed a lawsuit which alleged that ICE had made changes to the algorithm so as to eliminate entirely the recommendation that defendants be released.

Taken together, these examples only scratch the surface of the real challenges that emerge when advanced digital technologies meet migration management endeavours. These also apply to migration forced by climate change. Whatever the context, using advanced digital technologies in hostile or unstable situations has risks.

What would benefit future climate migrants more is consensus on the policies, programmes and procedures through which governments and powerful organisations will respond to them.

The world must agree on a clear definition of who is a climate migrant, what rights and protections they have and, in turn, what obligations and duties governments have to support them.

Advanced digital technologies, meanwhile, are short-term bandage ‘solutions’ that can pose severe risks to already vulnerable migrants, both now and many years after these same migrants have established new roots.

Update, 9 December 2022: Some information relating to the Immigration Policy Lab and Danish Refugee Council projects has been corrected and extended.