Abeba Birhane, named one of Time magazine’s 100 most influential people in AI in 2023, had been invited to speak at the opening ceremony of the AI for Good Summit, hosted this week in Geneva by the International Telecommunication Union (ITU). But two hours before her address, she was summoned to what she described as an “intense negotiation” during the rehearsal session with the event organisers, the Ethiopian-born researcher told Geneva Solutions.

Over the course of more than an hour, she says she was instructed to remove slides and change words referencing “Israel”, “Palestine” and “genocide” as well as any mention of technology companies, particularly Microsoft, Meta and Google in connection to war crimes or illegal activities.  A review of the original presentation by Geneva Solutions confirms that the version she used on the main stage at Palexpo is significantly different from the one initially submitted.

“I had uploaded my slides a week in advance,” said Birhane. “They had plenty of time to raise concerns. But instead, they waited until right before my keynote. It felt extremely stressful to experience censorship in real time,” she added.  Meredith Whittaker, the president of Signal, was in the room at the time of the incident and confirmed Birhane’s version of events.

The ITU didn’t confirm or deny what happened behind the scenes but in an email to Geneva Solutions explained that “all speakers are welcome to share their personal viewpoints about the role of technology in society”.

The annual summit has become the UN’s main forum for showcasing how artificial intelligence could be a force for good. But Birhane, who leads a team of researchers at the AI Accountability Lab at Trinity College Dublin and has served on the UN secretary general’s AI advisory body, does not share this view and came to Geneva to challenge it.

She had planned to use her keynote to question the growing use of “AI for social good” as a rhetorical shield by major tech firms. Her original presentation included publicly available documentation about Google and Microsoft’s contracts with the Israeli Ministry of Defence, as well as reports alleging that Meta trained its AI systems on pirated books. These examples, she argued, highlight the contradictions at the heart of the “AI for Good” premise. Microsoft has been a regular sponsor of the the event, while Mega has a booth to present its smart glasses.

“It’s a way of laundering accountability,” she said. “While companies claim to advance human rights and sustainability, they’re also supplying the technological infrastructure that is powering oppression, surveillance and, in some cases, atrocities.” Her comments come after a report last week by the UN special rapporteur on the rights situation in the Occupied Palestinian Territories, Francesca Albanese, named dozens of companies implicated in supporting Israeli settlements and in the war in Gaza.

The term “genocide,” Birhane was told, had to be replaced with the more diplomatic “war crimes.” One slide titled No AI for War Crimes drew particular concern from the organisers.

“It was where I drew the line,” she said. “I had already removed so much, but that slide was the core of my message. I refused to remove it.”

ITU told Geneva Solutions that it provided guidelines encouraging speakers to focus on meaningful issues and thoughtful, solution-oriented dialogue, and that all speakers were invited to rehearsal sessions to discuss their speeches with programme representatives prior to presentations.

“In the end,” wrote Birhane on BlueSky, “it was either remove everything that names names (big tech particularly) and remove logos or cancel the talk or turn it into a fireside chat without visuals.”

Researchers and NGO accusing the UN of making certain topics taboo has become more common in recent years. As reported by Geneva Solutions, UNFCCC, the UN climate body, faced accusations of censorship during Cop28 in the United Arab Emirates after it revoked the event badges of NGOs protesting over the situation in Palestine.

Birhane was the only speaker from an African country out of a long list of prominent tech entrepreneurs and political figures speaking at the opening event on Tuesday, including Amazon vice president Werner Vogels and Swiss federal councillor Guy Parmelin. She says her critical positions on AI as well as on the war in Gaza are well known. “I wish people would do their due diligence on my research and values before inviting me and getting surprised when they discover what I plan to talk about.”