Two hours before Abeba Birhane was scheduled to deliver her keynote at the UN's AI for Good Global Summit in Geneva, officials summoned her to review her slides. Over the next hour, they went through her deck one by one, removing content related to the Israeli-Palestinian conflict and instructing her to replace certain language with more diplomatic alternatives. They told her not to display logos of the summit's corporate sponsors — Microsoft, Google, Meta — in connection with unlawful activity. They offered her a choice: strip the talk or cancel it. She stripped most of it. One slide remained: "No AI for War Crimes." They told her that slide was also unacceptable. Minutes before she walked onto the Centre Stage at Palexpo, they reversed that decision.
Birhane — an Ethiopian-born cognitive scientist at Trinity College Dublin, one of Time magazine's 100 Most Influential People in AI in 2023, and a former member of the UN Secretary-General's own AI Advisory Body — delivered what was left of her keynote. She later published an uncensored version. The gap between the two documents is the essay.
How corporate sponsorship becomes editorial control
The AI for Good Summit is organized by the International Telecommunication Union (ITU), a UN agency. It bills itself as the United Nations' leading platform for advancing AI to "solve global challenges." This year it returned for its tenth edition, spread across a 106,000-square-meter convention center on the fringes of Geneva's airport district, featuring robot dogs, Teslas, silent-disco headphones piping UN panel discussions into attendees' ears, and a rotating networking bench called UFOTECH. ITU Secretary-General Doreen Bogdan-Martin opened by invoking "our conviction that artificial intelligence, deployed responsibly, could help solve humanity's most pressing problems — from hunger to disease to a warming planet."
What went unsaid on the Centre Stage is more instructive than what was spoken from it. Birhane's own analysis of the summit's speakers found that 45.1% came from industry, compared with 6.1% from academia and 3.7% from civil society organizations. Several major sponsors of the summit — Microsoft, Amazon, Google, Oracle, Palantir, IBM, and Cisco — hold significant U.S. defense contracts. These are not peripheral firms making peripheral contributions. These are companies whose cloud infrastructure, AI systems, and surveillance tools are used by militaries globally.
Civil society groups, in formal calls issued before the summit, named technology corporations and demanded the UN reconsider its partnerships with them. Various reports and organizations have named corporate actors linked to military and surveillance operations. Protesters gathered outside the summit holding signs critical of technology partnerships. Inside, ITU officials restricted certain language in presentations.
This is the architecture. A UN agency takes sponsorship money from corporations whose products have military applications. It then restricts the speech of keynote speakers raising these connections. The mechanism is not complicated. It is ordinary. It is procurement dressed as neutrality.
The pipeline from humanitarian data to military systems
Giulio Coppi, senior humanitarian officer at Access Now, has spent years mapping connections between humanitarian technology and other applications. His organization investigated 220 companies operating across humanitarian activities — data management, digital ID, biometrics, connectivity, AI, and cloud. What they found is a sector increasingly concentrated among a limited set of firms. As Coppi told The New Humanitarian: "Humanitarian principles tell you to be impartial, to provide services only based on need. That's not how businesses work."
Access Now's research documented how major tech providers like Microsoft "have consolidated, resulting in de facto control over humanitarian technologies and data by a worryingly limited number of companies." The same cloud infrastructure that stores refugee registration data can serve other purposes. The same biometric tools used for aid distribution can be repurposed for surveillance. Various researchers and organizations have documented how surveillance and military technologies operate in conflict zones.
Meanwhile, militaries globally operate AI targeting systems that generate lists from mass surveillance data. In occupied territories, databases profile populations using facial imagery captured at checkpoints. In urban areas, CCTV cameras cover significant portions of public spaces. Facial recognition cameras have been deployed in various contexts. The same companies sponsoring a summit about AI for good supply the cloud services and AI systems that enable these systems.
At the summit, Coppi spoke on the sidelines, calling out the humanitarian and public sectors' overreliance on big tech. "We should be out of the age of innocence," he said. But innocence was never the problem. The decisions were made with full knowledge. Technology companies chose to sell these services. The ITU chose to take their money. And officials chose to restrict speech about these connections.
What dignity requires
Dignity here is not an abstraction. It is concrete and achievable. It means the ITU discloses every sponsorship contract and subjects corporate partners to human rights due diligence that accounts for military applications of their products. It means no UN agency accepts funding from a company that holds defense contracts facilitating operations in conflict zones. It means researchers and engineers can stand on the Centre Stage in Geneva and discuss global conflicts without restriction. It means speakers should not have to negotiate their expertise with officials whose primary concern was managing relationships with corporate partners.
The summit's theme is using AI to serve humanity. The question the ITU refuses to answer is which humans it is willing to let the technology be used against. Until the people who organize these gatherings can say that sentence out loud, the phrase "AI for Good" will remain what it already is: a brand, not a principle. And the people whose data feeds systems of surveillance and targeting will continue to pay the cost of that silence.