At UN's AI Summit, Bold Visions Clash With Concerns Over Who Controls the Technology

The AI for Good summit, hosted by the UN's International Telecommunication Union in Geneva, assembled thousands of representatives from governments, tech firms, and civil society to explore how artificial intelligence might address humanity's biggest challenges—from disease and hunger to climate change. Yet beneath the optimistic framing lurked a more uncomfortable debate: whether the current trajectory of AI development would benefit the world broadly, or entrench existing power structures.

The Gap Between Aspiration and Reality#

Secretary-General of the ITU Doreen Bogdan-Martin opened the 10th annual conference with an uplifting message, insisting that "artificial intelligence, deployed responsibly, could help solve humanity's most pressing problems." But speakers and observers on the ground pushed back against what they saw as naive faith in the technology.

Humanitarian organizations working at the frontlines of global crises expressed particular frustration. Giulio Coppi, a senior officer at the campaign group Access Now, argued that the public and nonprofit sectors have grown dangerously dependent on large tech corporations, treating them "as your best friends" while signing opaque, multimillion-dollar contracts with public funds. "You can't even explain what's inside your tech stack, because it has kept changing," he cautioned.

The tensions erupted publicly when pro-Palestine activists interrupted a keynote by Amazon's chief technology officer, accusing the company of enabling Israeli military operations against Palestinians before being removed from the venue.

The Engineer's Problem: Defining "Good"#

A core challenge emerged repeatedly across sessions: how to actually build and deploy technology toward vaguely defined ends. As Harvard engineering professor Vijay Janapa Reddi pointed out, engineers need concrete specifications, not aspirational language. "When you're an engineer, good means nothing. I can't build you something that is good," he explained. The conference, for all its earnest discussions, struggled to translate ideals into verifiable technical requirements.

Access and Infrastructure as Geopolitical Stakes#

Much of the debate centered on compute power and who gets to use advanced AI models. Recent U.S. restrictions on exporting frontier AI chips, followed by reversals, and China's reported consideration of limiting overseas access to its models, underscored how the technology has become a development and geopolitical issue—not merely a technical one.