A new report from the United Nations warns that the rapid development of artificial intelligence could deepen global inequality, and that the opportunity for governments to establish effective oversight is shrinking. The findings come from a preliminary report by the UN's Independent International Scientific Panel on Artificial Intelligence, a body of 40 experts established by the General Assembly in 2025.
“The more AI advances without shared rules, the less say governments and people will have in the outcome,” said António Guterres, the UN secretary general, at a press conference on Wednesday. “Our message to governments is simple: do not wait … the science is here. We can no longer say we did not know what we do.”
Key Findings on AI Risks and Benefits
The report notes that generative AI can now write software, analyze vast datasets, produce lifelike images and video, and assist in scientific discovery. It also highlights the emergence of agentic AI, where systems complete complex tasks with minimal human involvement. The panel estimates that the difficulty of tasks these systems can handle has been roughly doubling every few months, raising concerns about monitoring and control.
Among the risks flagged are the use of AI to generate sexual abuse material and explicit deepfakes, disproportionately targeting women and children. The technology is also making disinformation more convincing and harder to detect, eroding public trust and democratic discourse. Cybersecurity threats are growing, with criminals using AI for fraud and social engineering, and the report warns of AI contributing to mental health crises, including suicide, through harmful interactions with vulnerable users. Additionally, the data centers powering AI are a growing source of greenhouse gas emissions.
However, the report also highlights benefits: AI models have mapped the structure of over 200 million proteins, accelerating drug discovery, vaccine research, and work on antibiotic resistance. The technology is helping flag food insecurity before it becomes a crisis and broadening access to education, mental health support, and tools for people with disabilities.
Uneven Global Distribution
The report underscores a stark imbalance in AI capacity. The United States controls roughly three-quarters of the computing capacity behind the world's leading AI supercomputers, with China holding around 15%. Together, they account for about 90% of that capacity, and the most advanced models are built by companies based in those two nations. Developing countries lack the talent, infrastructure, and funding to build or audit the AI systems they use. Without efforts to close this gap, the panel warns, AI risks widening global inequality.
This concentration of power has implications for Europe, where the EU has been working on its own AI governance frameworks, such as the AI Act. The report notes that more than 40 AI governance frameworks exist worldwide, but describes them as fragmented, inconsistent, and rarely tested for effectiveness. Much of the safety testing is still conducted by the same companies building the technology, raising questions about independence.
The panel calls for stronger third-party evaluation, more international coordination, and shared standards, alongside investment so countries can build the expertise and infrastructure needed to govern AI on their own terms. As the US lifts export controls on Anthropic's most advanced AI models, the urgency for a coordinated European response grows.
The report also highlights an "evidence dilemma": lawmakers need solid data before writing effective rules, but AI often evolves past that data before it is even compiled. This challenge is particularly acute for European regulators, who must balance innovation with protection of citizens' rights. The EU defence chief's warnings about budget cuts underscore the broader security implications of AI governance gaps.
As the window for control closes, the report serves as a stark reminder that without shared rules, the benefits of AI may be concentrated in a few hands, while the risks are borne globally. The panel's message is clear: governments must act now, before the technology outpaces their ability to govern it.
