By Fundación Maldita.es

In the morning of July 30th, the day already looked nothing like the slow-news Thursday one would expect when half the country was getting ready to go on vacation. The situation in Ceuta was already tense, but by noon our WhatsApp chatbot started exploding with messages: people were sharing videos of the border with us, eager to know if they were authentic—if that massive crowd of hundreds of people were actually crossing into Spain from Morocco, unimpeded, if dozens of them were drowning as they attempted to reach Ceuta by swimming.

Sure enough the crisis was real, not AI-generated, and unprecedented. At the newsroom we had vivid memories of another mass illegal crossing into Ceuta in 2021 and the wave of disinformation that ensued, but soon we realized the scale was just uncomparable. Back then 10,000 people crossed the border irregularly and this time, according to government figures, over 70,000 would make it into Ceuta—a city of 84,000—and at least  141 people would die trying.

It was too big. We could not just spot the inevitable disinformation stemming from the crisis and debunk—this time we would need to understand how the crisis was possible in the first place, why none saw it coming at that scale, which actors were pushing a particular explanation or what happened, and specially: could it happen again in a few days? That would require much more from us than past crises: more insights, more depth, more answers.

Technology and the human factor

By Friday morning it was too clear that part of the team at Fundación Maldita.es would work nonstop through the weekend and in the coming weeks, but there were things we just could not do ourselves. Our monitoring systems are extremely effective when it comes to disinformation in Spanish because we have a very committed community that shares tips with us through different channels on a daily basis, because we complete those insights with social media listening and our expert knowledge, and finally because we have developed the technology that organizes all that information to separate signal from noise so we can prioritize what matters the most. But regarding the situation in Ceuta, the Spanish language was just one side of the story.

If we had the ambition to get the full picture, we needed to look at the other side of the conversation too, the one happening mainly in Moroccan Arabic and other sub-Saharan languages. We had to build a system from scratch that would allow us to understand the information (and disinformation) that the people who had crossed into Ceuta had as they made their decision, or that others in Morocco were consuming as they pondered whether to do the same.

Our monitoring and tech teams quickly joined forces to make sure an AI agent was processing those conversations happening on social media or in groups where migrants organized. But as we already knew full well, having access to the information is not the same as assessing it properly. There was a lot of context that our Madrid-based team could not fully understand, so that is when we decided to join forces with our colleagues at the Arab Fact-Checkers Network, whose Moroccan fact-checkers would prove invaluable to understand what was really going on.

With some emergency funding from the Facteur EU-supported project, we were able to pay for that collaboration and the inordinate amount of work that would allow us to start finding the meaningful answers we were looking for. For example: how the mass crossing was organized on Facebook groups and the ongoing plans to do it again two weeks later, partly inspired by disinformation about “an open border” and an Instagram algorithm that really loved the idea and promoted it through algorithmic amplification. We also documented those suspicious job offers to work in the west popping up online in Morocco and the particular reading some were promoting about an Spanish judicial ruling on the mechanism to expel those swimming into Ceuta as a major cause for the crisis.

Nothing of that would have been possible if we would have taken the Ceuta crisis as “business as usual”, but again the facts on the ground were anything but usual. Nevertheless, as we investigated all those angles in depth we still had to ramp up our fact-checking about disinformation in Spanish. A month after the mass crossing, we have already debunked 55 individual claims about what is happening in Ceuta and analyzed the involvement of foreign actors in pushing disinformation about the crisis.

When the next big crisis hits

At Fundación Maldita.es we see collaboration as central to our mission: we are part of the European Fact-Checking Standards Network (EFCSN) as are more than 60 other fact-checking organizations, we are also in the Spanish-Portuguese EDMO hub Iberifier, and we are signatories of the Code of Practice on Disinformation, an official code of conduct under the EU Digital Services Act for the most important social media companies. But as much as we are committed to collaborate with others—particularly during crises—so people can see the context they need when seeing disinformation online, the Ceuta crisis was also unprecedented in this regard.

It was clear from the start that disinformation in Moroccan Arabic was putting lives at risk, by manipulating some of the migrants and giving them false information as they made their decision to cross. On the other hand, the mechanisms of the Code of Conduct and the DSA are designed for crises affecting Europeans directly and for disinformation circulating in European languages. Our own agreement with Meta to participate in its third-party fact-checking program only covers Spanish-language content and accounts that pertain to Spain. Nevertheless, in coalition with other Spanish fact-checkers, we decided to explore a conversation with platforms that could be effective in the challenge at hand.

X and YouTube are not interested in collaborating with independent fact-checkers, but even with platforms such as Facebook, Instagram, and TikTok—that have some collaboration with fact-checkers in Europe under the Code—it was unclear how to proceed. The code’s signatories task force operates a ‘rapid response system’ that is activated during national elections to flag harmful disinformation so platforms can assess whether it violates its policies or the applicable national law, but that model was not consistent with a situation in which disinformation in Arabic was affecting Moroccans in Morocco.

We settled on the creation of a time-limited information-exchange mechanism that would allow us to share information with those platforms about disinformation contents, narratives, and other harmful misleading patterns happening on their services. The collaboration helped prevent some migrants from being taken advantage of, but it also underlined some previous concerns we had about this kind of mechanism.

When a crisis hits, it is already too late to start designing such collaborations. Platforms need to have systems in place that have clear definitions, work flows, and built-in traceability and accountability guarantees so organizations such as ours can know exactly how to contribute and what to expect of these mechanisms. And this particular crisis highlights the need for all platforms to build collaboration with local fact-checkers in countries where a crisis might arise, which is to say in all countries. The reason we could be effective at understanding the full picture of what was happening in Ceuta was because we were lucky to count on our partners for the Arab Fact-Checkers Network at ARIJ who had the invaluable local knowledge that no technological solution can fully substitute for.

Before the next big crisis comes—and we can be sure it will come—we need to have serious conversations so we are better prepared, and we can be more effective at preventing disinformation harms and building collaborations that matter for the people most affected by it.