Table of Contents
Key insights
The report summarizes the findings in seven key areas:
- AI science, advances and trajectories
- Social applications in science, health, education and agriculture
- Economic impact
- Impact on safety, systems and the environment
- Human rights, information and democracy
- Cultural advantages, autonomy and child safety
- Management, governance and reliability
“The science is there,” said UN Secretary-General António Guterres when presenting the report. “We can no longer say we didn’t know. What we do with it is now up to all of us.”
The further AI advances without common rules, the less influence governments and people have on the outcome, the UN chief said, adding: “My message to governments is simple: don’t wait.”
With the aim of creating shared understanding and evidence at this critical juncture, the Preliminary report of the independent international scientific panel on AI: Evidence-based assessment of the opportunities, risks and impacts of AI was written by the first global, fully independent scientific body dedicated to assessing its real impact on economies and societies.
Read the full report Here.
Why it matters
Globally, over a billion people are now using conversational AI every week, while governments make momentous decisions in the face of great uncertainty, with rapidly changing, often contradictory sources of evidence and perspectives that do not necessarily reflect local realities.
“Well used, AI could be the most powerful driver of development, accelerating global progress in everything from health and hunger to learning and climate” said the UN chief, “but the body is equally clear about the harm that artificial intelligence can cause.”
In fact, as AI’s capabilities grow, so do the stakes – the central challenge that the committee wants to address.
Read our AI explainer here.
Better world or catastrophic damage?
The panel, made up of 40 leading scientists and experts from all regions, outlines AI trends and warns that current protective measures cannot keep up, said its co-chair Yoshua Bengio.
“AI capabilities exceed both scientific understanding and the adaptability of governments“Given mounting evidence of fraudulent AI behavior, science cannot currently guarantee that AI will not cause catastrophic harm as its capabilities increase, either alone or through malicious users.”
To act effectively, he said, global policymakers need to understand these systems, and the panel provides just that: a rigorous, common scientific foundation “that guides our shared path forward.”
Key findings
Detecting breast cancer earlier, accelerating vaccine development, and improving healthcare services are just a few groundbreaking AI achievements, but limitations and challenges remain, including:
- AI adoption has broadly accelerated, although in different countries and sectors
- Access and use vary widely, with adoption in the Global South lagging far behind the Global North
- There are significant differences in computing infrastructure and computing models between advanced economies, reflecting existing inequalities
In addition, development is highly concentrated, with recent estimates suggesting that the United States accounts for 75 percent of the computing power among the world’s top 500 AI supercomputers while China accounts for 15 percent, and that companies in both countries are developing almost all of the leading general-purpose models.
Understand risks
Understanding and managing AI risks is critical, the report says. The panel’s co-chair, Maria Ressa, added that the risks to societies, security and the human species were already “too high.”
“The technology is transformative, but if the world continues on this path, humanity will not realize the promised benefits,” she warned.
Here are some of the panel’s warnings:
- There are no scientific guarantees that AI agent systems will not violate instructions, and evidence is accumulating of cases where this is already the case
- AI agent systems will soon complete tasks that currently take human programmers days or weeks to complete, but their deployment raises pressing questions for labor markets, cybersecurity and the controllability of future AI systems
- Sycophantic AI behavior, in which reactions reinforce users’ existing beliefs regardless of accuracy, has been linked to several serious psychological incidents, including documented deaths
- Criminals and malicious actors have been documented to use AI systems to support cyberattacks
- Advanced technical capabilities can enable inexperienced private actors to use AI in malicious ways in a range of applications such as fraud and disinformation
- There is a lack of reliable methods for maintaining control of highly autonomous AI systems
Many of these harms disproportionately affect already disadvantaged populations, the report says.
Unlock benefits, reduce risks
To minimize AI risks and benefit from this technological tool, good governance is required.
There are concrete next steps to address current gaps, but each requires sustained investment in Member States’ capacities to design, evaluate and deploy AI.
Seizing these opportunities safely requires targeted investments and policies to promote equitable access and reward innovation while preventing exploitation of the vulnerable.
“AI will not close divisions”
Dozens of different governance tools aimed at embedding ethics and human rights in AI systems are already in use in various jurisdictions, but they are fragmented, concentrated in a few companies and rarely measure effectiveness in practice, the report says.
Amandeep Gill, Under-Secretary-General and Special Envoy for Digital and Emerging Technologies, said the new report provides a common scientific language that decision-makers can now use in the future.
“AI will not close divides by itself,” he said.
The benefits come where institutions, skills and data already exist, and where they don’t, the same technology can displace workers, increase inequality and make communities dependent on systems designed without them in mind, he explained.
“These realities are now on record, independently verified and can no longer be ignored.”
The report’s findings will be presented to governments at the first UN Global Dialogue on AI Governance, taking place in Geneva on July 6 and 7.
Watch the report release and press conference here:
https://news.un.org/en/story/2026/07/1167853
