On July 1, 2026, the UN's Independent International Scientific Panel on AI released its preliminary report. The document emphasizes that, if deployed thoughtfully, AI could serve as a powerful engine for development, accelerating progress in healthcare, food security, education, and climate action.
Appointed by the UN General Assembly in February, this group of 40 scientists and experts produced the first independent, global scientific analysis of the technology. Among the contributors is Professor Vukosi Marivate of the University of Pretoria. The report has been distributed to all governments and is available to the general public.
The authors candidly address the downside as well. Rapid, unchecked scaling carries significant risks, ranging from harming users' mental health and use for destructive purposes to broad impacts on social, economic, and environmental systems. Reliable methods for maintaining control over highly autonomous AI systems do not yet exist.
Particular attention is paid to the concentration of power. The United States accounts for 75% of the computing resources in the largest AI clusters. This could lead to authoritarian capture and the erosion of democratic accountability. AI capabilities are already outstripping scientific understanding and governments' ability to adapt.
What else does the report highlight?
1. Deceptive AI Behavior: Panel co-chair Yoshua Bengio stated that scientists are recording increasing instances of deceptive behavior in advanced AI models. He explicitly noted that science cannot currently guarantee that AI will not cause catastrophic damage through autonomous actions or malicious use.
2. Language Inequality and Medical Errors: The report highlights a stark language gap, noting that current AI models support only a tiny fraction of the world’s 7,000+ languages. Machine translation errors in low-resource languages are already leading to fatal consequences, such as incorrect medical diagnoses and treatment plans.
3. Specific Threats: The document outlines specific threats ranging from deepfakes to biological risks.
4. The Evidence Dilemma: The complexity of tasks AI handles doubles every few months. Despite the existence of over 40 different regulatory frameworks worldwide, they remain fragmented and largely untested in practice.
5. Transition to Agentic Systems and Technology Convergence: The panel predicts a rapid shift toward autonomous "agentic AI" systems capable of executing multi-step tasks independently. In the long term, AI is expected to merge with other transformative technologies like quantum computing and biotechnology.
6. Environmental Transparency and the Secretary-General's Initiative: Amid the report's release, UN Secretary-General António Guterres is championing an "AI Environmental Transparency Initiative." This requires companies to fully disclose data on carbon footprints, water consumption, and land use for data centers, aiming to transition the industry to renewable energy by 2030.
7. The Idea of an "IAEA for AI": Ahead of the Geneva Dialogue, there is growing support for creating a specialized international agency for AI management, modeled after the International Atomic Energy Agency (IAEA), specifically for technologies impacting national security.
8. Risk of Digital Colonialism: Highlighting the risk of "digital colonialism," the report notes that while over a billion people use conversational AI weekly, developing nations often lack the technical expertise to evaluate advanced models. Consequently, they risk becoming mere "rule-takers" with no voice in policy creation, worsening global inequality.
The report will form the basis for discussions at the inaugural Global Dialogue on AI Governance in Geneva on July 6–7. The forum will bring together all 193 UN member states along with representatives from industry, academia, and civil society.
