A SCIENTIFIC panel created by the United Nations released its first independent assessment of artificial intelligence (AI) this month, and the warning at its center is a blunt one. The technology is moving faster than governments can measure it, let alone manage it. Behind the report is the Independent International Scientific Panel on Artificial Intelligence, a body of 40 independent experts that counts one Filipino among them: Maria Ressa, who co-chairs the panel with Canadian computer scientist Yoshua Bengio. They presented the findings in New York on July 1.
At the launch, UN Secretary-General Antonio Guterres said the panel is intended “to help the world separate fact from fakes, and science from slop.”
The UN General Assembly created the panel through Resolution 79/325 in 2025, but its 40 experts, drawn from all five UN regional groups, were appointed only in February 2026. That left the group three months from its first meeting in March to produce a report. But the mandate stops at the evidence: the panel documents what the science says and where experts disagree, and governments decide what to do with it. As Bengio put it, “making recommendations is not its mission.”
Speed without guardrails
What does the evidence say? AI capabilities, tracked through benchmark tests, have roughly doubled their rate of improvement since April 2024. Over a billion people now use conversational AI every week. And the industry behind it has become highly concentrated. Of the computing power in the world’s 500 largest AI supercomputers, 75 percent is in the United States and 15 percent is in China, leaving the rest of the world to divide the remaining 10 percent. US institutions produced 59 notable AI models in 2025, China produced 35, and the rest of the world produced 13 combined. And 91 percent of those models came from private companies rather than governments or universities.
“When forty scientists from forty different contexts have to agree, you don’t drift toward the most alarming claim. You build toward the center,” Ressa said. “What you are receiving is the floor of our concern. Not the ceiling.”
Where PH fits
That concentration is where the report gets close to home. It counts 118 countries, most of them in the Global South, that are not engaged in major AI governance discussions. Fewer than a third of developing countries have a national AI strategy on paper. Then there are the languages. The world speaks more than 7,000 of them, but AI systems are trained and tested on only a small fraction. The panel points to Tigrinya as an example. In that language, spoken by millions in Eritrea and northern Ethiopia, machine translation rendered “smallpox” as “syphilis” and changed intravenous antibiotics into intravenous insecticides. Mistakes like that can kill. Any country still building its own AI capacity should pay attention, ours included.
Ressa was clear that this is no attack on the technology. “The technology is extraordinary. The opportunities are real — in science, in education, in health care, in agriculture,” she said.
The report itself credits AlphaFold with predicting the structures of more than 200 million proteins, now used by more than 3 million researchers, and documents gains in agriculture, disease screening and education, but with a catch: those gains appeared where AI was grounded in local languages and paired with trained teachers and health workers. Access to AI tools does not equal benefit.
When machines don’t listen
The harder findings concern AI agents and mental health. There is no scientific guarantee that autonomous AI systems will follow their own safety instructions, and laboratory cases exist where they have not. Closer to daily life, AI chatbot use for therapy or companionship now reaches at least 24 percent of American adults, ahead of any safety consensus, and court cases allege that chatbot responses to users expressing suicidal thoughts have contributed to deaths.
Which brings me back to Ressa. Filipino readers have watched her fight over what counts as true online and who gets to decide.
“A lie told a million times becomes a fact. I lived that,” she said. “Without facts, you cannot have truth. Without truth, you cannot have trust.”
I got a brief word with her and asked what her decade of warning about social media taught her about governance losing the race again with AI.
“Short answer if we fail again: it could be catastrophic,” she told me. “You know I’ve said this over and over: without information integrity, we lose every battle — climate change, development, health ... it starts with reclaiming lost ground from the days of social media.”
On July 6 and 7, the panel carried that evidence to Geneva for the UN’s first Global Dialogue on AI Governance, where all 193 member states had a seat. Ressa urged ministers and heads of state not to “wait for certainty that will not arrive in time to matter.” Guterres called for an AI Child Safety Pledge and for every major AI company to power every data center with renewable energy by 2030.
For those 118 countries, the more useful question sits closer to home: Who in each capital picks up the pen next?
