Artificial intelligence is advancing faster than governments can write rules to manage it. That is the central finding of a preliminary report from the UN’s Independent International Scientific Panel on Artificial Intelligence, which published its first findings ahead of the UN Global Dialogue on AI Governance, scheduled to begin in Geneva on July 6. The event will bring together member states to discuss how the world should manage a technology that is reshaping nearly every industry at once.
The panel, made up of scientific experts from around the world, is not a regulatory body. Its job is to assess the technology and give policymakers the information they need to act. But its first report makes clear that the gap between what AI can do and what oversight systems can handle is already wide, and growing.
The reason governance is falling behind, according to the panel, is structural. Authorities typically need a solid base of scientific evidence before introducing regulations. By the time that evidence exists for a given AI capability, the technology has often already moved on. The report notes that the complexity of tasks AI models can handle has been doubling every few months, a pace that no traditional regulatory process was designed to match.
The report is not entirely negative on AI. It acknowledges real benefits that are already playing out in the world, including:
- Faster drug discovery and vaccine development
- Contributions to antibiotic resistance research
- Early detection of illnesses like breast cancer
- AI-powered early warning systems for food insecurity
These are not hypothetical gains. Researchers and health systems are already using AI tools in these ways, which makes the governance question more urgent, not less. Getting the rules wrong could slow down the benefits while leaving the harms unchecked.
And the harms are significant. The report documents several categories of risk that AI systems have either already created or are likely to create:
- Generation and distribution of nonconsensual sexually explicit deepfakes, including child sexual abuse material
- Spread of false information that looks credible
- AI tools that help criminals carry out cyberattacks
- Sycophantic models that reinforce harmful behaviors in users, in some cases linked to suicide risk
- Increasingly autonomous systems that are harder to monitor and control
- Environmental and community harm from large data center buildouts
The deepfake issue alone has already triggered government action. California opened an investigation into Grok earlier this year over nonconsensual deepfakes and child sexual abuse material. The UN report puts that case in a broader context: this is not an isolated platform problem, it is a systemic one.
The panel is also direct about inequality. AI development is concentrated in a small number of countries, primarily the US and China. Most developing nations lack the infrastructure and technical expertise to benefit from these tools in any meaningful way. Without deliberate effort to change that, the report warns, AI could make existing global inequalities worse. The panel writes that without proper safeguards, AI technologies “could deepen inequality, spread misinformation, threaten human rights, disrupt labor markets” and become a tool controlled by very few governments and companies.
What the panel is calling for is not a slowdown in AI development. It is calling for stronger independent evaluation of AI systems, greater international cooperation, and common standards that would make AI more transparent and accountable across borders. These are not new ideas, but the report gives them fresh urgency by anchoring them in a concrete assessment of where the technology is right now.
The panel is expected to publish a more comprehensive report next year. For now, its preliminary findings set the terms for the Geneva dialogue. Whether member states can agree on anything meaningful remains to be seen, but the scientific case for acting quickly is now on record.
