Forty of the world’s top scientists and experts are sounding both an optimistic and cautionary note about artificial intelligence, releasing the first-ever independent global scientific assessment of the technology on behalf of a United Nations panel.
The preliminary report is set to be delivered to government representatives at the inaugural UN Global Dialogue on AI governance, scheduled for July 6 and 7 in Geneva. A more complete and comprehensive version of the report is expected to be released next year.
Panel members were selected from every region of the globe and serve three-year terms without ties to any government, institution, or private company.
Among the report’s key findings:
Policymakers need solid scientific evidence to properly regulate AI, but the technology is advancing faster than researchers can fully understand it — and faster than governments can keep up. Very few tools currently exist to control highly autonomous AI systems.
Panel co-chair Yoshua Bengio pointed to growing evidence that AI systems are capable of deceptive behavior. He said science cannot promise that AI won’t cause catastrophic harm, in his words, “either on its own or due to malicious users,” as its capabilities continue to grow.
The report stated plainly: “The potential benefits of AI are enormous. The rapid, unchecked deployment of the technology at scale also presents considerable risks, including harms to the mental health of users, potential use as a destructive tool, impacts on social, economic and environmental systems, and challenges associated with controlling the technology.”
While AI use has grown rapidly across the world, that growth has been uneven. More than one billion people now use conversational AI on a weekly basis, but uptake in developing nations trails far behind wealthier countries.
The development of AI itself is even more concentrated geographically. The United States accounts for 75% of the computing power among the world’s 500 most powerful AI supercomputers, while China holds 15%.
The report also flagged a significant language gap: while more than 7,000 languages are spoken around the world, current AI models are trained on only a small fraction of them. Machine translation of some languages contains frequent errors — errors that can directly affect health diagnoses and treatment decisions.
On the risk side, the report cited growing concerns about AI’s impact on human rights, social systems, and the environment. It noted that AI-generated child sexual abuse material and deepfake-enabled sexual violence are appearing more frequently online.
The panel also warned that AI makes it far easier to produce and distribute persuasive content at massive scale, leading to what the report called a “gradual erosion of information integrity that can weaken public trust, social cohesion and democratic deliberation.”
Finally, the report noted that most countries — including advanced economies — do not have the technical expertise needed to evaluate the most powerful new AI systems or to play a meaningful role in shaping how they are governed.
