President Donald Trump’s energy secretary has downplayed the threat of rising temperatures at a fossil fuel industry-funded event in London, even as millions in the U.K., France, Spain and Italy endure a deadly record-breaking heatwave.
In his remarks, Wright lauded energy from fossil fuels as good for the economy and society at large.
“Always more people die in the winter than die in summer,” Wright said, “because cold is a vastly larger killer than heat is,” citing the excess deaths across Europe due to high energy prices after Russia’s 2022 invasion of Ukraine.
Wright also urged the U.K. to “change course” and embrace fossil fuels, and called climate change a “slow-moving phenomenon” that would be addressed by technology.
Wright also urged the U.K. to “change course” and embrace fossil fuels.
Experts have said for years that while technology can play a part in lowering carbon dioxide levels in the atmosphere, it cannot replace ending the use of fossil fuels.
Reform UK leader Nigel Farage and Conservative Party leader Kemi Badenoch were also on the speaker schedule for the three-day ARC conference, which wraps up Thursday. Both back scrapping the U.K.’s landmark Climate Change Act, and have called for new North Sea oil extraction.
The U.K. is currently seeing record-breaking temperatures as part of a heat wave across Europe. At least 15 people died last month in the U.K. in water-related incidents during hot weather, and the national weather service has issued an extreme heat warning for June, when temperatures were expected to top 100 degrees.
Multiple news outlets are reporting that since Thursday, at least 40 people have drowned in France while seeking escape from the searing heat. The New York Times reported that temperatures may approach 110 degrees in central France.
Experts agree that heat waves are made more likely by the climate crisis. Before joining Trump’s Cabinet in 2025, Wright was CEO of the hydraulic fracturing (fracking) services company Liberty Energy.
The Trump administration is the most anti-climate and pro-fossil fuel government in recent U.S. history, spreading climate denial, blocking renewable energy, and boosting oil and gas extraction.
After his ARC address, Wright was interviewed by Matt Ridley, a Tory peer and newspaper columnist who is also an adviser to the Global Warming Policy Foundation, a U.K. climate denial think tank.
Ridley made reference to the U.K. heat wave, saying: “It’s hot in London today, Chris. I am speaking to you from a continent where air conditioning is a sin against Gaia.”
He went on to ask Wright about the benefits of fracking in the U.K. “Ten years ago, you and I tried to persuade the British government to embrace shale gas”, he said. “If we had succeeded, Britain could have been self-sufficient in gas now, couldn’t it?”
Wright replied that a “shale gas revolution” in the U.K., where fracking — the environmentally damaging process of extracting oil and gas from rock by injecting it with water, gravel and chemicals — is currently banned, could have led to an “industrial renaissance” that would have cut electricity bills and created jobs for “blue-collar workers”.
Ridley also asked whether the “battle for the commons” on climate change had been won in the United States. Wright replied: “I will say we are winning. We haven’t won fully yet, but we’re winning by leaning in.”
He later added: “Understand climate change for what it is: a slow-moving phenomenon that ultimately will be addressed by better technologies.”
“I will say we are winning. We haven’t won fully yet, but we’re winning.”
This reflects Wright’s long-standing views. In a video posted to LinkedIn in January 2023, Wright said, “There is no climate crisis”.
The United Nations Intergovernmental Panel on Climate Change, the world’s leading climate science body, has said it is “unequivocal” that human influence has caused “unprecedented” global warming, and that this is increasing the threat of floods, drought and heat waves.
Wright’s comment about hot weather echoed remarks earlier at the ARC event by Bjorn Lomborg, a Danish climate crisis denier. During his slide presentation, Lomborg said that “extreme heat almost everywhere on the planet is the smallest death issue. … Remember many more die from extreme cold.” Wright and Lomborg are known to be friends.
Wright’s involvement with ARC goes back to 2023, when he spoke at the group’s first conference in his capacity as an oil executive.
Wright also appeared by video link at ARC’s 2025 conference in London, this time as Trump’s energy chief, vowing to “get out of the way” of coal, oil and gas, and calling the U.K.’s 2050 net zero target “a sinister goal” that would “impoverish” people.
ARC CEO Philippa Stroud, another Tory peer, and Wright confirmed this week that they talked on the phone about the ARC conference ahead of his Cabinet nomination in 2025.
The red emergency light is flashing on America’s democracy dashboard, like a damaged aircraft teetering toward a mountain. Elon Musk becoming the planet’s first trillionaire should make us tremble for the future of self-governing republics. It’s as if we’re bringing back modern pharaohs to dominate our societies.
The concern here isn’t with wealth per se. It’s the tremendous power of concentrated wealth to distort markets, politics and society. When you have Musk’s level of wealth, you’re no longer just buying another mansion or private jet (of which he already has several). You’re buying a media outlet, a senator, and maybe, in the case of Musk, elevating a president.
Musk has no inhibitions about deploying the power of his considerable wealth. He bought Twitter, one of the public squares of our time, and transformed it into X, a partisan disinformation platform rife with hate speech and extremism.
Musk has no inhibitions about deploying the power of his considerable wealth.
In the 2024 election cycle, he donated $291 million to President Donald Trump and Republican candidates, according to Open Secrets. As Michael Mechanic wrote in Mother Jones, “Musk expended 0.1 percent of his wealth in the process and got far more in return.” Mechanic notes, “The Trump administration promptly shelved dozens of investigations into Musk’s companies.”
More dire still, DOGE cuts to the U.S. Agency for International Development and other humanitarian aid programs have contributed to an estimated 750,000 lost lives. The projected deaths from these cuts run into the millions.
Musk was further rewarded with lucrative government contracts for SpaceX, Starlink and other companies he controls. In early 2025, The New York Times reported on a boost in multibillion-dollar contracts for Musk’s companies as the Trump administration took power.
That was Musk as a “mere” centi-billionaire. What other power might Musk be able to wield as the world’s first trillionaire?
But it’s not just Musk. America’s 16 centi-billionaires (including Musk) have a combined wealth of $4 trillion. And the 977 billionaires on the Forbes U.S. wealth list now own a combined $9.24 trillion, according to analysis by Americans for Tax Fairness.
This isn’t a partisan concern. Whether it’s liberals like George Soros and Tom Steyer or right-wingers like Musk and Peter Thiel, this concentration of power and influence should trigger the flashing red light. It’s never a good thing for anyone to have the power of modern-day pharaohs. Musk was the top political donor in 2024, but five other billionaire households gave over $100 million to candidates.
America’s 16 centi-billionaires (including Musk) have a combined wealth of $4 trillion.
Billionaires — and soon trillionaires as well — are spending hundreds of millions of dollars to influence our elections while working Americans struggle to afford food, housing and healthcare. It’s clearer than ever that those two facts are connected. We need to get serious about curbing this billionaire influence and supporting regular people — starting with a wealth tax.
Oxfam observes Musk could give $100 to every person on Earth and remain one of the world’s 10 richest people. A 10% wealth tax on Musk’s fortune alone, the international aid alliance estimates, could end global extreme poverty and lift 800 million people above the extreme poverty line. Imagine the revenue and investment possibilities of a global wealth tax on all billionaires.
The planet’s first trillionaire is not a sign of economic health. It’s an indicator of extreme inequality and the dangers of concentrated power.
Kinshasa, République démocratique du Congo – Dans les collines poussiéreuses de Mongbwalu, cité minière de la province de l’Ituri, personne ne parlait encore d’Ebola lorsque les premiers décès ont commencé à se multiplier. Tout a commencé fin avril, avec le retour d’un infirmier malade à Bunia, chef-lieu de l’Ituri province avec ses quelque 1,5 millions d’habitants. Il meurt quelques jours plus tard.
Comme le veut la tradition dans de nombreuses communautés de la région, parents, voisins et proches se rassemblent alors dans la ville voisine de Mongbwalu pour rendre un dernier hommage au défunt. Son épouse, qui l’avait soigné et avait participé aux rites funéraires, décède peu de temps après. Plusieurs personnes manipulent les deux corps au cours des cérémonies. En l’espace de deux semaines, au sein de cette même famille, 15 personnes vont décéder à Mongbwalu. À ce moment-là, personne ne soupçonne encore la présence du virus Ebola.
« Beaucoup pensaient qu’il s’agissait de fétiches ou d’un phénomène surnaturel. Personne n’imaginait qu’il pouvait s’agir d’Ebola », raconte à Truthdig Isaiah Katavu, habitant de Bunia.
« Personne n’imaginait qu’il pouvait s’agir d’Ebola. »
Ces six dernières semaines, la lutte contre l’épidémie d’Ebola en RDC s’est heurtée à de nouveaux obstacles : désinformation, pénurie de matériel médical, conséquences des coupes budgétaires de l’USAID et extrême pauvreté, des millions de personnes étant confrontées à une famine sévère. Le personnel soignant a également témoigné auprès de Truthdig des difficultés rencontrées pour transporter équipes et équipements dans un contexte de contrôle territorial morcelé, de conflit armé et de massacres.
Mais la clé de la lutte contre l’épidémie réside dans le dépistage rapide des personnes exposées au virus, explique à Truthdig le professeur Jean-Jacques Muyembe, co-découvreur du virus Ebola en 1976 ainsi qu’un traitement par anticorps et qui est aujourd’hui directeur général de l’Institut national de recherches biomédicales de Kinshasa. Mais, ajoute-t-il, les divers défis économiques, sanitaires et liés aux conflits rendent cette tâche extrêmement difficile.
Une détection tardive
En Ituri, bon nombre attribuent d’abord les décès à des pratiques et forces mystiques, et le virus peut ainsi se propager sans être détecté. Les morts s’accumulent. Les familles enterrent leurs proches sans savoir précisément ce qui les a emportés.
Pendant ce temps, les autorités sanitaires peinent à identifier l’origine de la maladie. Treize échantillons sont finalement envoyés à Kinshasa pour analyse. Plus de soixante personnes auraient déjà perdu la vie lorsque les résultats confirment, le 15 mai, le retour d’Ebola dans cette partie du pays.
Ce retard a mis en lumière des problèmes plus profonds au sein du système de surveillance mis en place pour détecter le virus. Le test de diagnostic, GeneXpert, détecte une souche différente, et s’est donc révélé négatif. Une rupture de la chaîne d’approvisionnement a empêché le maintien au froid des échantillons destinés aux analyses. Divers responsables, notamment des professionnels de santé et des responsables politiques, n’ont pas donné l’alerte à temps. Les réductions de l’aide humanitaire ont également nui à la surveillance.
Des agents de santé s’occupent d’un patient atteint d’Ebola au centre de traitement de Rwampara, en Ituri (Congo), le jeudi 18 juin 2026. (Photo AP/Moses Sawasawa)
Le 18 mai, de nouveaux prélèvements confirment d’autres contaminations. Parmi les personnes infectées figure également Peter Stafford, un citoyen américain travaillant depuis 2023 à l’hôpital de Nyankunde, à une quarantaine de kilomètres de Bunia. Après plusieurs semaines de traitement, il est déclaré guéri le 6 juin dans un établissement hospitalier à Berlin.
L’épidémie de maladie causée par le virus Ebola, qui provoque une fièvre hémorragique extrêmement contagieuse, poursuit sa progression en RDC. Au 21 juin, le ministère de la santé a déclaré 956 cas et 247 morts confirmées – sans compter les 19 cas répertoriés en Uganda.
Le 14 juin a enregistré une hausse record du nombre de cas, et endiguer l’épidémie s’avère difficile pour diverses raisons. Cette flambée présente une particularité majeure : elle est causée par le variant Bundibugyo et non par la souche Zaïre, responsable de plusieurs épidémies antérieures. À ce jour, aucun vaccin homologué ni traitement spécifique n’existe contre cette variante.
L’Ituri, principal foyer de l’épidémie, constitue qui plus est l’une des régions les plus complexes du pays. Province minière riche en or, elle est traversée depuis des années par des conflits armés, des déplacements massifs de populations et une profonde crise humanitaire. La province enregistre plus de 900 000 déplacés victimes des groupes armés.
Une riposte compromise par la désinformation
Contenir Ebola en Ituri est également devenu une lutte contre la désinformation. Des publications virales sur les réseaux sociaux affirment qu’il n’y a pas d’Ebola dans la région, et on estime qu’une personne sur trois en Ituri ne croit absolument pas en l’existence du virus.
La méfiance historique à l’égard des autorités sanitaires est renforcée par l’insécurité, les tensions politiques, et les inégalités, ainsi que par les inquiétudes concernant le trafic d’organes et la promotion de traitements non validés.
Les informations fiables sont rares et les résidents dépendent du bouche à oreille pour les dernières nouvelles, ce qui peut déformer les faits. Certains habitants voient Ebola comme un stratagème destiné à attirer des financements internationaux.
« Dans plusieurs villages reculés, beaucoup pensent qu’Ebola est un business », explique Isaiah Katavu. « Certains disent que ce sont des microbes apportés par les Blancs pour les injecter aux Africains. »
La méfiance autour de la 17ᵉ épidémie d’Ebola en Ituri s’explique par plusieurs facteurs
notament les mauvaises expériences du pays au cours des épidémies précédentes. La théorie de l’“Ebola business” remonte à l’épidémie de 2020, lorsque trois ambassadeurs de pays ayant envoyé de l’aide financière à la RDC se sont érigés contre la corruption. Des listes de paie gonflées désignaient alors quelque 4,000 agents supposément affectés à la réponse contre Ebola pour lutter contre seulement 120 cas.
Selon un rapport du groupe d’Etude sur le Congo, des groupes armés ont monnayé la violence, se faisant « acheter » par la Riposte et n’hésitant pas à prolonger l’épidémie pour continuer à tirer profit de la crise.
« La plus grande faiblesse dans la riposte reste le manque d’adhésion communautaire. »
Cette méfiance a parfois dégénéré en violence. Le 21 mai, à l’hôpital général de référence de Rwampara, en périphérie de Bunia, une foule en colère s’en est prise à un centre d’isolement destiné aux patients atteints d’Ebola. Les manifestants contestaient les circonstances du décès d’un proche et exigeaient la restitution du corps. La situation a rapidement dégénéré, et la police est intervenue, procédant à des tirs de sommation. Deux tentes d’isolement ont été incendiées, et plusieurs patients ont pris la fuite.
« La plus grande faiblesse dans la riposte reste le manque d’adhésion communautaire », explique à Truthdig Augustin Bedidjo, coordonnateur de l’Association des Exploitants Miniers Artisanaux pour la Pacification et la Reconstruction de l’Ituri.
« Beaucoup de familles continuent à se méfier des équipes sanitaires, » dit-il. « Certaines cachent même des malades ou refusent de signaler des cas suspects. »
Chaque malade caché complique le travail de traçage des contacts, ajoute-t-il. « Lorsqu’une famille cache un malade, les équipes ne peuvent pas identifier rapidement les personnes exposées. Chaque retard augmente le risque de propagation. »
Plusieurs familles dans de nombreux villages continuent à observer leurs rites, tout en vivant dans la précarité suite au manque d’eau, à la promiscuité dans les habitations, et à la nécessité de travailler tous les jours pour subvenir à leurs besoins. Ils peinent ainsi à respecter les mesures sanitaires de base comme le lavage des mains, la limitation des contacts avec les malades, et le respect des enterrements sécurisés.
« Nous faisons face à plusieurs obstacles, » poursuit Bedidjo : « le déficit d’adhésion communautaire, la vulnérabilité économique des populations et, surtout, le manque de financement des organisations locales ».
La vulnérabilité économique du pays est aggravée par les carences du système de santé. Ces deux facteurs illustrent un paradoxe amer en RDC, l’un des pays les plus riches en minéraux au monde mais qui figure pourtant parmi les dix pays les plus pauvres. Malgré ses ressources minières inexploitées estimées à 24 000 milliards de dollars, la plupart des compagnies minières en RDC sont étrangères, et ces richesses ne restent pas dans le pays, dont la population doit tenter de survivre avec moins de 3 dollars par jour.
Des agents de santé se désinfectent après avoir préparé le corps d’une victime d’Ebola à la clinique Citadelle de Bunia, au Congo, le vendredi 12 juin 2026. (Photo AP/Moses Sawasawa)
Les agents de santé traitant Ebola en RDC affirment qu’ils manquent de tentes d’isolement individuelles pour les patients, ainsi que d’équipements de protection suffisants pour les travailleurs. Actuellement, les tentes d’isolement pour plusieurs personnes débordent, et il n’y a aucun lit disponible dans les hôpitaux de la région touchée.
Selon Bedidjo, la décision de Donald Trump de couper le financement de l’USAID début 2025 a eu de fortes répercussions au sein des organisations gouvernementales et non gouvernementales en RDC, fragilisant encore plus les capacités locales de sensibilisation et de surveillance.
En 2024, le financement américain à la RDC s’élevait à 1,4 milliard de dollars. En 2026, il avait chuté à 146 millions de dollars. Les programmes destinés à détecter les cas d’Ebola, à alerter les communautés en cas d’infection et à distribuer des kits d’intervention ont donc vu leur financement réduit. La réduction des financements a contraint les organisations humanitaires à diminuer leurs effectifs, tout en maintenant leurs opérations avec un personnel réduit.
« Ce sont elles qui peuvent convaincre les familles, expliquer les mesures sanitaires dans les langues locales et réduire la méfiance, » explique Bididjo. « Mais sans soutien financier ni logistique, leur capacité d’action reste limitée », ajoute-t–il.
Les conflits armés entravent la riposte à Ebola.
Comme lors des précédentes flambées, l’insécurité demeure l’un des plus grands défis de la riposte. Certaines parties de l’Ituri, proches des zones touchées, sont encore en proie à la violence armée. Le 4 juin, quatre personnes ont été tuées dans le village de Tchelo, du district de Djugo, en Ituri, lors d’une attaque attribuée à des miliciens de la Coopérative pour le développement du Congo (CODECO) – une coalition de groupes armés issus de la communauté lendu. Très actif dans la province d’Ituri, riche en minéraux, le mouvement est régulièrement accusé d’attaques contre des civils et des sites miniers.
Dans le territoire de Mambasa (toujours en Ituri), où certains cas ont été confirmés, les rebelles des Forces démocratiques alliées (ADF), liés au groupe État islamique, poursuivent également leurs incursions meurtrières. Leur plus récente incursion a eu lieu le 31 mai, causant la mort de 21 personnes en une nuit.
Pour les professionnels de santé, ces attaques compliquent considérablement la surveillance épidémiologique.
« Lorsqu’il y a une attaque, les populations fuient dans toutes les directions », explique à Truthdig le docteur Louis Mutuza, médecin basé à Beni et acteur de la riposte Ebola entre 2018 et 2020. « On perd alors la trace de certaines personnes qui peuvent être porteuses du virus. »
Dans un tel contexte, identifier les contacts et suivre les chaînes de transmission devient un défi . « Cette épidémie est plus complexe que les précédentes parce qu’il existe aujourd’hui une multitude d’acteurs armés sur le terrain », explique Mutuza.
« Lorsqu’il y a une attaque, les populations fuient dans toutes les directions. »
« Certaines zones rendent compte aux autorités gouvernementales, » ajoute-t-il, « tandis que d’autres sont sous contrôle de groupes armés. Cela complique énormément la coordination. »
Dans plusieurs parties des provinces voisines du Nord-Kivu et du Sud-Kivu, l’autorité de l’État demeure limitée. Certaines zones de ces provinces, notamment autour de Goma et de Bukavu, sont contrôlées de facto par des groupes armés. Cette fragmentation du territoire ralentit le déploiement des équipes médicales et perturbe les opérations logistiques.
« Travailler dans ces zones est un travail à très haut risque », insiste Mutuza. « Chaque matin, nous partons sans savoir si nous rentrerons le soir. Les équipes médicales peuvent être attaquées, kidnappées, ou se retrouver au milieu d’affrontements. »
Pour Mutuza et ses collègues qui ont participé à la précédente riposte entre 2018 et 2020, les souvenirs de cette épidémie servent de rappel de ce qui pourrait mal tourner aujourd’hui. Mutuza se remémore que, lors des épidémies antérieures, certains déplacements nécessitaient de longues négociations avec des groupes armés. Parfois, les équipes sanitaires devaient expliquer leur mission pendant plusieurs jours avant d’obtenir l’autorisation d’accéder à certaines localités.
« Dans les zones les plus dangereuses, les opérations étaient parfois menées sous escorte de la Mission de l’Organisation des Nations Unies pour la stabilisation en République Démocratique du Congo (MONUSCO), ou des forces armées congolaises, » ajoute-t-il. Malgré ces précautions, certaines régions demeuraient pratiquement inaccessibles.
« Il existait des endroits où aucune escorte ne pouvait entrer » ajoute-t-il. « Dans ces cas-là, il fallait respecter des règles imposées localement pour être accepté. »
L’essentiel est de détecter tous les cas
En réponse à l’épidémie, le Rwanda et l’Ouganda ont fermé leurs frontières, et l’aéroport de Goma reste paralysé. Le ravitaillement humanitaire est gravement perturbé par le manque de vols et de personnel dans cette région désormais classée en zone rouge, se désole le docteur.
« Le grand défi de cette riposte sera de détecter tous les contacts et d’isoler rapidement ceux qui développent des symptômes »,explique le microbiologiste Muyembe. Et note que l’expérience des épidémies précédentes a démontré jusqu’où certains patients seraient prêts à aller pour éviter les autorités.
« Certains patients se réfugiaient auprès de groupes armés pour éviter les équipes de santé, » explique-t-il.
« Nous sommes généralement parvenus à dialoguer avec ces groupes, » ajoute Muyembe. « Nous leur expliquions que s’ils gardaient les malades auprès d’eux, l’épidémie finirait également par les atteindre. »
Cette approche avait permis d’ouvrir des couloirs humanitaires et de poursuivre les activités de surveillance dans des zones autrement inaccessibles. « Nous appliquerons la même méthode aujourd’hui », assure-t-il.
La RDC compte un médecin pour 5 000 habitants (contre un médecin pour 350 habitants en Angleterre, par exemple). Malgré les difficultés considérables, ces professionnels de santé sont pleinement mobilisés dans la lutte contre l’épidémie.
KINSHASA, DRC — When people first began dying in the dusty hills of Mongbwalu, a gold-mining town in the Democratic Republic of Congo’s Ituri province, few suspected Ebola. It all started in late April, when a nurse returned to Bunia, the 1.5-million people capital of Ituri. He fell ill and died a few days later.
As is tradition in many communities in the region, relatives, neighbors and loved ones then gathered in nearby Mungbwalu to pay their last respects. The nurse’s wife, who had cared for him and taken part in the funeral rites, became sick and died shortly after. Several other mourners came into direct contact with his and her bodies during the ceremonies. Within two weeks, 15 people from one family alone died in Mongbwalu. Still, no one suspected the Ebola virus.
“Many thought fetishes or a supernatural phenomenon were involved. No one imagined it could be Ebola,” said Bunia resident Isaiah Katavu.
“No one imagined it could be Ebola.”
Now, over the past six weeks, the battle to contain Congo’s Ebola outbreak has only faced more obstacles — from misinformation, to limited health supplies and the impact of cuts to the U.S. Agency for International Development, to extreme poverty, with millions in the area facing severe hunger. Healthcare workers also describe the precariousness of transporting teams and equipment amidst divided territorial control, armed conflict and massacres.
But the key to fighting the epidemic is quickly finding those exposed to the virus, said professor Jean-Jacques Muyembe, a microbiologist who co-discovered the Ebola virus in 1976 as well as an antibody treatment and who now serves as director general to the National Institute of Biomedical Research in Kinshasa. But, he adds, the various economic, health and conflict challenges are making that very difficult.
Late detection
In Ituri, the attribution of the Ebola deaths to mystical practices and forces allowed the virus to spread undetected. The death toll rose as families buried loved ones without knowing what had killed them.
Health authorities, meanwhile, struggled to identify the source of the illness. Thirteen tissue and body fluid samples were eventually sent to Kinshasa for analysis, but it wasn’t until May 15 that the results confirmed the return of Ebola to the east of the country. By then, more than 60 people had already been reported dead.
The delay exposed deeper issues in the system of surveillance in place to detect the virus. The diagnostic test, GeneXpert, detects a different Ebola strain than the one now circulating, and had come back negative. A disruption in the supply chain prevented the refrigeration of samples intended for analysis. Various officials, including healthcare professionals and politicians, did not raise the alarm. Reductions in humanitarian aid also hampered surveillance work.
On May 18, further samples confirmed additional infections. Among those was Peter Stafford, a U.S. citizen who had been working at the Nyankunde hospital, about 40 kilometers from Bunia, since 2023. On June 6, after several weeks of care, doctors at a Berlin hospital declared him cured of the virus.
Health workers tend to an Ebola patient at the Rwampara Treatment Center in Ituri, Congo, on June 18, 2026. (AP Photo/Moses Sawasawa)
Ebola, which provokes a highly contagious hemorrhagic fever, is still spreading. As of June 21, the country’s health ministry has reported 956 cases and 247 confirmed deaths, and there are also 19 confirmed cases in Uganda.
June 14 saw a record-breaking increase in cases, and stopping the epidemic is proving difficult for a variety of reasons. The current outbreak differs from previous ones because it is caused by the Bundibugyo strain, not by the Zaire strain responsible for several previous outbreaks. There is currently no approved vaccine or targeted treatment for the Bundibugyo strain.
Ituri province, at the center of the outbreak, is one of the most complex regions in the DRC. Rich in gold deposits, it has been scarred for years by armed conflict, mass displacement and a humanitarian crisis. Violence by armed groups has displaced a total of more than 900,000 people across the province.
Misinformation undermines a coordinated response
Containing Ebola in Ituri has also become a struggle against misinformation. Viral social media posts are claiming there is no Ebola in the region, and an estimated 1 in 3 people in Ituri don’t believe Ebola exists at all.
Historical distrust in health authorities is augmented by the generally unsafe environment, political tensions and inequality, as well as by concerns about organ trafficking and the promotion of unvalidated treatments.
Reliable information is rare and residents depend on word of mouth for updates, which can distort the facts. Some residents see Ebola as a ploy designed to attract international funding.
“In several remote villages, many see Ebola as a business,” Katavu says. “Some say that these are microbes brought by white people to inject into Africans.”
Horrible past experiences also underlie the mistrust around this 17th Ebola outbreak in Ituri. The “Ebola business” belief dates back to the 2020 outbreak, when three ambassadors of countries providing public aid denounced corruption in Congo. Inflated payroll lists showed 4,000 staff members were reportedly assigned to the Ebola response to deal with some 120 contaminations.
According to a report by the Groupe d’Etude sur le Congo (Congo Research Group), armed groups also monetized violence. Some were found to have been bought off by the Riposte — Congo’s political, institutional, infrastructural and financial assemblage responding to the outbreak — and to have prolonged the epidemic in order to continue to profit from the crisis.
“The biggest weakness of the response remains the lack of community support.”
Such mistrust has sometimes degenerated into violence. On May 21, at the Rwampara Treatment Center on the outskirts of Bunia, people attacked an isolation center for Ebola patients. Protesters contested the circumstances of a relative’s death and demanded the body be returned. The situation quickly escalated, with police intervening and firing warning shots. Two isolation tents were set on fire and several aid workers fled.
“The biggest weakness of the response remains the lack of community support,” Augustin Bedidjo, coordinator of the Association of Artisanal Miners for the Pacification and Reconstruction of Ituri, tells Truthdig.
“Many families still distrust health teams,” Bedidjo says. “Some even conceal sick relatives, or refuse to report suspected cases.”
Every unreported case makes contact tracing more difficult, he explains. “When a family hides a sick person, teams cannot quickly identify those who have been exposed. Every delay increases the risk of transmission.”
Families across several villages have continued observing their funeral rituals while also living in precarious conditions due to water shortages, overcrowding in their homes and the need to work every day in order to survive. As a result, they struggle to observe basic health precautions such as hand-washing, limiting contact with the sick and ensuring safe burial practices.
“We face several obstacles,” Bedidjo adds, “A lack of community support, the economic vulnerability of the population and above all, the lack of funding for local organizations.”
Health workers disinfect themselves after preparing the body of an Ebola victim at Citadelle Clinic in Bunia, Congo, on June 12, 2026. (AP Photo/Moses Sawasawa)
The economic vulnerability is compounded by healthcare shortfalls. Both of these factors are hallmarks of a bitter paradox in Congo, which is one of the world’s richest mineral-producing countries, yet ranks among the top 10 poorest countries. Although the country has untapped mineral resources estimated at $24 trillion, most mining companies in Congo are foreign, and much of that wealth never reaches ordinary Congolese people, who must try to survive on less than $3 per day. (Congo has one doctor on average per 5,000 people; England, by comparison, has one for every 350 people.)
Healthcare workers treating Ebola in Congo say they lack individual isolation tents for patients as well as sufficient protective gear for workers. Currently, multiple-person isolation tents are overflowing, and there are no available beds in hospitals in the affected region.
Further, Bedidjo says U.S. President Donald Trump’s decision to cut USAID funding in early 2025 has had a significant impact on governmental and nongovernmental organizations in the country, further weakening on-the-ground capacity for awareness-raising and monitoring.
In 2024, the U.S. sent $1.4 billion in aid to Congo. By 2026, it had fallen to $146 million. As a result, programs designed to detect Ebola cases, warn communities about new infections and distribute response kits have seen their funding slashed. Humanitarian organizations — often the groups with the best access to local communities — have been forced to reduce staff while attempting to maintain operations.
“They are the ones who can convince families, explain health measures in local languages and reduce mistrust,” Bedidjo says. “But without financial or logistical support, their ability to act is limited,” he added.
Armed conflict inhibits the Ebola response
As with previous outbreaks, the threat of violence remains one of the biggest challenges for the response. Parts of Ituri near the affected areas are still plagued by bloodshed. On June 4, four people were killed in the village of Tchelo, in Djugu district, during an attack attributed to militiamen from the Cooperative for the Development of Congo (CODECO). CODECO is a rebel network of Lendu fighters. Active in the resource-rich Ituri province, the group is regularly accused of attacks against civilians and mining sites.
Members of the Allied Democratic Forces rebel group, linked to the Islamic State, have also kept up their deadly incursions in the Mambasa Territory, also in Ituri, where Ebola cases have been confirmed. Their most recent incursion took place on May 31 and reportedly killed 21 people in one night.
For healthcare professionals, these attacks significantly complicate epidemiological surveillance.
“When there is an attack, people flee in all directions,” said Louis Mutuza, a physician based in Beni, just south of Bunia, who participated in the Ebola Riposte between 2018 and 2020. “We then lose track of people who may be carrying the virus.”
In this fraught context, identifying contacts and tracing transmission chains becomes a challenge. “This epidemic is more complex than previous ones because there are now a multitude of armed actors on the ground,” Mutuza says.
“When there is an attack, people flee in all directions.”
“Some areas report to government authorities,” he adds, “while others are under the control of armed groups. This greatly complicates coordination.”
In several parts of the nearby provinces of North Kivu and South Kivu, state authority remains limited. Some areas of those provinces, notably around the cities of Goma and Bukavu, are de facto controlled by armed groups. This fragmentation of territorial control and governance slows down the deployment of medical teams and disrupts logistical operations.
“Working in these areas is a very high-risk job,” Mutuza says. “Every morning, we leave without knowing if we will return in the evening. Medical teams can be attacked, kidnapped or find themselves in the middle of clashes.”
For Mutuza and his colleagues who participated in the 2018-20 Riposte, memories of that epidemic are reminders of what can go wrong. He recalls how moving around in some areas required lengthy negotiations with armed groups. Sometimes, health teams had to explain their mission for several days before being granted permission to access certain areas.
“In the most dangerous areas, operations were sometimes conducted under the escort of the United Nations Organization Stabilization Mission in the Democratic Republic of the Congo or the Congolese forces,” Mutuza says. However, despite these precautions, some regions remained virtually inaccessible.
“But there were places where no escort could enter,” he adds. “In those cases, you had to follow locally imposed rules to be accepted.”
The key is detecting all cases
In response to the epidemic, Rwanda and Uganda have closed their borders, and the nearby Goma airport remains paralyzed. According to Mutuza, humanitarian aid deliveries have been severely disrupted by the lack of flights and personnel in the region, now classified as a “red zone” because of security concerns.
“The main challenge of this response will be to detect all contacts and quickly isolate those who develop symptoms,” says Muyembe, the microbiologist, noting that past experience demonstrates the lengths to which some patients would go to avoid authorities.
“Some even sought refuge with armed groups to avoid healthcare teams,” he says.
“We were generally able to engage in dialogue with these groups,” Muyembe adds. “We explained to them that if they kept the sick with them, the epidemic would eventually reach them as well.”
This approach made it possible to open humanitarian corridors and to continue monitoring activities in otherwise inaccessible areas. “We will apply the same method today,” he insists.
Between February 2022 and September 2025, Bellingcat staff and volunteers collected, geolocated, and shared more than 2,500 incidents of civilian harm following Russia’s full-scale invasion of Ukraine.
As part of this effort, Bellingcat tested a new machine learning model intended to rank Telegram social media posts on their likelihood of containing incidents of civilian harm.
This novel methodology dramatically reduced the search and selection time required, freeing researchers to focus on verifying incidents of civilian harm – not just searching for them.
This piece documents our methodology, ethical considerations and lessons learned in the hope that others researching similar topics can benefit from our work.
Open source research into civilian harm is still a relatively new field and it presents many challenges – one of the biggest is organising and sorting through the huge volume of user generated content being produced to find what is relevant.
Machine learning, a form of artificial intelligence that uses algorithms to identify patterns from large amounts of data and make predictions, can make this task more efficient.
With ongoing conflicts involving large amounts of civilian harm occurring in Sudan, and much of the Middle East, this guide aims to offer those covering these conflicts an example of how machine learning can be used to help find and sort incidents. You can also access the Code Notebook for our model here.
We defined “civilian harm” not just as civilian deaths or injuries resulting from armed conflict, but also the broader and delayed effects on civilians from mental trauma, loss of livelihood, displacement, destruction of infrastructure and more. This definition was informed by the Protection of Civilians bookon civilian harm.
Initial Telegram Dataset
Each Telegram post containing civilian harm which had already been manually verified by researchers was used to build an initial dataset of confirmed cases of civilian harm, which data scientists call positive instances. We collected a total of 5,848 unique URLs for these Telegram posts. For our manual collection we reviewed posts on relevant Telegram channels, working through oldest to newest posts each day. Assuming that a given post made it to our geolocated incidents list, it meant the researcher who flagged it also looked at the posts that appeared before and after it on Telegram and did not flag those ones, so we selected the 10 posts surrounding the verified civilian harm post as our additional dataset of posts that did not contain civilian harm. After excluding any deleted or duplicate posts, we ended up with 48,545 non-civilian harm posts, our negative instances.
The choice to overrepresent negative instances aims at better reflecting the real world and increasing data available for model training.
We enriched each URL with metadata from the Telegram API, such as the time of publication, reactions or textual content. As some of these posts had been deleted, we completed the missing data points with previously preserved versions from our Auto Archiver database, only available for the positive instances.
Feature Engineering
Training a machine learning model requires numerical data, as these models compute a prediction score based on mathematical operations.
We built these by converting raw data from our initial dataset, such as keywords signalling potential civilian harm, into numerical scores (or “features”) that the model could interpret, with the aim of increasing the model’s ability to identify patterns. This process, known as feature engineering, can significantly improve model results because it allows data scientists to suggest explicit context knowledge.
A full list of features we used to train the model can be found in the code notebook accompanying this piece. Many features were directly inspired by researchers’ input from their experiences manually screening cases of civilian harm by sorting through a set number of Telegram channels and inspecting each post individually.
Several of the features used were directly built from the metadata contained in each Telegram post including media_type, day_of_week; or binary ones: forwarded, edited andreply_to.
Other features included engagement information: views, forwards, total_reactions, and even individual features for most used emojis including the reaction_crying_face to count emoji.
Converting Text to Numbers
To embed the experience from the manual collection process, researchers put together a list of keywords both in Ukrainian and Russian that, to them, signalled posts likely to show civilian harm. For instance, “Шахед” and “КАБ” translated to “Shahed” and “Guided aerial bomb” respectively. We created a numerical feature to count their frequency.
In addition, we included several generic English-language keywords which meaningfully signalled potential civilian harm, such as “injured”, “school affected” and “hospital affected” that were only used for generating semantic similarity scores.
A semantic similarity score is a calculation used to determine the proximity in meaning between different words and phrases. To get the semantic similarity between the post text and each of our keywords, we represented each in a list of numbers via a Sentence Transformer model, which converts words into numerical representations called vectors that a computer can understand.
We then calculated the level of similarity between each vector using cosine similarity, one of the most popular methods for measuring similarity between two pieces of text.
Due to how embeddings work, this calculation results in a figure on a scale from -1 (no semantic proximity) to 1 (same meaning). For example, the words “hurt” and “injured” would have a high similarity score, while “residential” and “injured” would have a negative score as the words are not semantically similar.
Finally, to enable the model to identify the relevance of each post to civilian harm in Ukraine, we used a multilingual text transformer from the BERT family of language models to represent the entire post’s text as a vector of 768 numerical values. This model can efficiently represent text from many languages in a way that captures meaning: the same sentence in different languages will generate similar embeddings, and trained machine learning models can detect patterns in the embeddings.
It is important to note that for this initial prototype of a civilian harm detection model, we did not include any features derived from media content such as photos and videos, although that would be a logical next step in attempting to improve model performance.
Selecting, Training and Evaluating Models
With 54,393 rows of 893 numerical features each, we selected four machine learning algorithms to train our predictive models.
We chose Logistic Regression as a baseline algorithm due to its simplicity. We also selected three other “best in class” models, Random Forest, XGBoost, and LightGBM. These choices centred on the interpretability of the models and their ability to work on tabular data of this size. For example, we avoided neural networks due to a lack of interpretability and because those models work best with a larger dataset.
To genuinely assess the performance of the trained models, we split our dataset into three parts:
A training set – the data the models were trained on (60 percent of the full dataset’s rows)
A validation set – used for an intermediary evaluation when tuning model parameters (20 percent of all rows)
A test set – hidden for the final performance assessment, so the models were evaluated on unseen data (remaining 20 percent of rows)
We used a stratified split to divide the dataset instead of a random split. This method ensured the proportion of positive instances (i.e. confirmed cases of civilian harm) remained consistent across all three sets at about 11 percent.
To measure the performance of machine learning models, we ran them through the test set and measured the number of correct and incorrect predictions. Models output a likelihood between 0 and 1 that each Telegram post contains civilian harm, and we tried to find a cut-off threshold that leads to a good balance between flagging almost every post (0.1) or flagging very few (0.9).
There are two main types of evaluation metrics to gauge a model’s prediction power. Recall asserts what fraction of positive instances (i.e. known civilian harm posts) were correctly flagged as such. Precision measures the fraction of posts flagged as civilian harm that are indeed civilian harm posts.
During the training phase, we tuned the models to maximise average precision (PR-AUC), a metric that summarises precision across all recall levels. While this method also accounts for precision, it prioritises recall, which is preferable for this use case as it steers model selection to reduce the number of civilian harm posts that are skipped.
The following table sorts models from best to worst PR-AUC against a baseline of a coin-flip predictor. ROC-AUC and F1 are two other evaluation metrics included as sanity checks. Simply put, ROC-AUC measures the probability of ranking two instances, one negative and one positive, correctly; F1 balances precision and recall equally and its best cut-off threshold value.
Model test scores comparison, XGBoost stands out in every relevant metric evaluated.
From these results, we selected XGBoost as our final model as it had the best scores when compared across all metrics.
Interpreting the Model
Because these models are interpretable, we can understand which features are the most useful when predicting whether a post includes civilian harm. The above table shows the top 10 features that most strongly signal the XGBoost model to make a decision:
semantic_keywords_similarity: the semantic proximity between the post text and manually selected keywords “casualties”, “damage” and “civilian harm”
bert: the model was able to discern meaning from the text with the same strength as some of the other features in this list – there are three cases of this in the top 10
reaction_crying_face: reactions with crying face emojis on the post
group_of_messages: whether a post contains multiple media files
keywords_in_text: the number of custom Ukrainian or Russian keywords in the post
These results generally tally with what you might expect when selecting Telegram posts for instances of civilian harm, including that posts that generate a lot of emotional engagement and posts using keywords about civilian harm were among those most likely to contain content related to this topic. Not all models had the same top features as XGBoost. In fact, for the Random Forest model the most important feature was the number of crying face emojis present in a post, a soft pattern highlighted by researchers when this methodology was first imagined.
LLM Results and Comparison
Retroactively, we decided to run a sample of the same test dataset through different large language models (LLMs) to gauge their ability to make these same predictions.
We aimed to include an LLM-generated score as an extra feature for our trained models, which would be captured as relevant if it correlated with the correct predictions.
To start, we selected two local models, the 1B and 4B variants of Gemma 3 from Google DeepMind, and two cloud-hosted models, Gemini 2.5 flash and Gemini 3.5 flash. With this selection, we hoped to compare results across a wide range of models’ expected performance.
We generated a 400-row stratified sample (preserving the same proportion of real civilian harm instances) from the test dataset used for the custom models. For each of the four LLM models, we ran two tests: one where only the Telegram post message was sent, and another including both the message and the engineered features (excluding the text embeddings, as the model had direct access to the text). In the prompt for each model, we asked for a score between 0 and 1. We then evaluated the results as we did for the custom models.
The above table shows that LLMs can indeed extract value from the engineered features. All four LLMs surpassed the baseline Logistic Regression model in our tests, yet none of them performed better than the other custom-trained models, and XGBoost remained the one with the highest PR-AUC.
Still, Gemini 2.5 Flash performed better than its newer version 3.5 and even achieved a slightly higher best F1 score than any other model. While this is a good result, for the flagging of civilian harm posts, the PR-AUC remains the crucial metric, as it captures the model’s ability to identify infrequent instances of civilian harm while minimising false positives.
Ethical Considerations
Introducing an instrument of automated decision-making into a process of detecting civilian harm brings inherent ethical questions. These include automation bias, or how humans tend to blindly place faith in machine-generated recommendations; algorithmic bias, or how the results of these models echo the same patterns present in the training data, including under- or over-representation of types of civilian harm.
The decision to test an automated methodology for this particular project came from the fact that there were limited resources for both steps in the process – the detection of potential civilian harm and its actual verification. Historically, we built an enormous backlog of unverified incidents because a lot of time had to be spent on monitoring the most recent events so that potential evidence would be captured and preserved as soon as possible.
The automation of this process also reduced the exposure of researchers to a significant amount of unpleasant and distressing visual and text content, reducing the burden of exposure to traumatic content.
For this project, we tried to ameliorate the ethical challenges with a number of strategies including randomly flagging posts not captured by any model, monitoring which features models relied on to make decisions, and by doing historical comparisons of patterns in data.
Additionally, as stated above, for this initial prototype of a civilian harm detection model we did not include any features derived from the media content itself. In the future, it would be a logical next step in attempting to improve the model performance, to include the media from the posts – but using AI to review actual media comes with additional ethical challenges such as model bias.
Because of the opaque ownership of many LLM companies and their generative nature, the use of LLMs for an extra feature presented additional ethical challenges including privacy and safety concerns considering the sensitive nature of the data. Our model did not rely on LLMs, though we retroactively ran a sample through it.
How the Model Fits into the Bigger Picture
After selecting this model, we created a user interface where researchers could view a list of Telegram posts sorted from most to least likely to contain indications of civilian harm. The user interface was designed for quick triage and integration, where a positive confirmation from researchers would instantly send the post to the Auto Archiver (Bellingcat’s tool for preserving digital content) and then transfer it to ATLOS (our internal collaborative verification platform). Bellingcat staff and volunteers could then manually verify incidents. Researcher input was constantly stored so that this data could be used to improve the model in the future.
Preliminary feedback indicated that the AI model was useful. Not only were we able to reduce time and harm from scouring through dozens of war reporting Telegram channels, researchers also reported that the stream of new posts being added to the verification backlog were capturing real and diverse cases of civilian harm.
Despite the focus on civilian harm and Telegram (highly popular in Ukraine and Russia), this pipeline is generic and can be adapted to other conflict monitoring tasks. How easily this can be done does depend on how open the social media platform is and whether it is possible to scrape posts from it. Apart from that, it is easy to incorporate new features and data, and cheap to automatically retrain, test and deploy models as the system receives more human input.
Looking forward, sorting through overwhelming amounts of data in a conflict will continue to be challenging. Hopefully, this methodology can help newsrooms, conflict monitoring organisations, and others find the balance between ethical considerations and resources in order to carry out open source investigations on civilian harm and human rights violations.
Bellingcat is a non-profit and the ability to carry out our work is dependent on the kind support of individual donors. If you would like to support our work, you can do so here. You can also subscribe to our Patreon channel here. Subscribe to our Newsletter and follow us on Bluesky here, Instagram here, Reddit here and YouTube here.
Philippine authorities arrested a senior member of Japan’s “Luffy Group,” a transnational crime syndicate accused of running large-scale fraud operations targeting Japanese citizens. The suspect, wanted under a Tokyo court warrant, was detained in Makati City and is facing deportation proceedings, according to local authorities.
Investigators said the group operated sophisticated scams by impersonating Japanese police officers and finance ministry officials to gain victims’ trust. Once victims allowed supposed officials into their homes, members of the syndicate allegedly distracted them, secretly swapped their ATM cards with counterfeit ones, and later used the stolen cards to withdraw money. Police said parts of the operation were run from the Philippines and warned other members may still be active in the country.
Since Tuesday, United Nations peacekeepers have detected no airstrikes or new missile fire in southern Lebanon. The lull remains fragile, however, as Israeli drones continue to fly over the area and military operations continue on the ground.
Two deadly earthquakes struck Venezuela less than a minute apart on Wednesday, with magnitudes of 7.2 and 7.5, causing at least 164 deaths and widespread destruction in and around the capital, Caracas as international assistance begins to arrive early Thursday, with UN agencies rapidly deploying aid, support and rescue teams. UN News app users can follow here.
UN teams scrambled on Thursday in support of the international response to the devastating double earthquake disaster in Venezuela, where buildings lie flattened and people are likely still trapped in the capital, Caracas, and beyond.
emoji.
