UN’s first global AI science panel warns the window to govern the technology is closing

Artificial intelligence is advancing faster than governments can regulate it, and the world’s first global scientific body on the technology says the moment to act is now. That is the conclusion of the preliminary report from the UN Independent International Scientific Panel on Artificial Intelligence, launched on Wednesday ahead of a major governance summit in Geneva. The panel warns that the window to establish effective global oversight remains open, but may not stay that way for much longer. Its central worry is concentration. According to the report, the United States controls around three-quarters of the computing power behind the world’s leading AI supercomputers, while China holds roughly 15%. Together that gives the two countries about 90% of the compute that trains the most capable systems. Most of the frontier models themselves are also built by companies headquartered in those two nations. Speed is the problem The report describes an “evidence dilemma” at the heart of AI policy. Policymakers want reliable scientific data before they legislate, the panel notes, but by the time enough evidence accumulates, the technology has often moved on. Researchers cited in the report say the complexity of tasks these systems can complete has been doubling every few months. The next wave is already arriving in the form of AI “agents” that can plan tasks, use digital tools, and write software with little human oversight. That autonomy is spreading fast into finance and commerce, where firms are already handing agents real decisions and, in some cases, money to spend. The report is careful not to read as a doom document. It points to AI systems that have predicted the structures of more than 200 million proteins, work led by Google DeepMind’s AlphaFold and now used to speed up drug discovery and research into antibiotic resistance. Doctors are using AI to catch diseases such as breast cancer earlier, and early-warning systems are flagging food insecurity before it tips into crisis, according to the panel. Used responsibly, the panel argues, the technology could accelerate progress towards the UN’s Sustainable Development Goals across health, education, and agriculture. The risks it wants governments to watch The same tools are fuelling sexual abuse material and deepfakes, with women and children most at risk, the report finds. It also flags AI-generated disinformation, cyberattacks, and fraud, alongside the mental-health harms of systems that can reinforce dangerous beliefs. Then there is the physical footprint, since the energy-hungry data centres that power AI are adding to greenhouse gas emissions, a bill the UN has separately urged firms to come clean on. Poorer nations, the panel warns, risk being locked out entirely, unable to build, inspect, or audit the systems they increasingly depend on. Established by the UN General Assembly in 2025, the body brings together 40 experts from every region, serving in their personal capacity. Its role is scientific rather than regulatory, meaning it assesses evidence and publishes reports that governments can draw on, without setting rules or enforcing standards. The report notes that more than 40 AI governance frameworks and ethical guidelines already exist worldwide, yet remain fragmented, inconsistent, and rarely tested to see whether they work. Many safety assessments, it adds, are still run by the very companies building the technology. The panel calls for stronger independent evaluation, international cooperation, and common standards, an argument that echoes the direction set by the EU’s AI Act. Its findings now feed into the UN Global Dialogue on AI Governance, which opens in Geneva on 6 July. The panel’s bottom line is blunt: AI is neither inherently good nor bad, and its impact will hinge on choices made today. Get the TNW newsletter Get the most important tech news in your inbox each week.