This is the International Report on Security in Artificial Intelligence 2026, prepared with the guidance of more than 100 experts from over 30 countries and international organizations, including the European Union (EU), the Organization for Economic Cooperation and Development (OECD), and the United Nations (UN).

Economic (OECD) and the United Nations (UN).

According to the study, the United Arab Emirates and Singapore have the highest adoption rates, with more than half of the working-age population using AI tools.

Meanwhile, the economies with the highest adoption rates are in Europe and North America.

These estimates are based on anonymous data, mainly from Microsoft Windows users, adjusted to take into account different rates of personal computer ownership between countries and their use on mobile devices.

The document presents a comprehensive analysis of the progress, capabilities, and dangers associated with advanced AI.

Prepared by a global panel of specialists, the material highlights how improved reasoning and autonomy capabilities have facilitated accelerated, albeit geographically uneven, adoption.

Critical risks such as the generation of malicious content, the manipulation of beliefs, potential cyberattacks, and biological threats are examined.

The text also evaluates the effectiveness of various technical mitigation measures, warning that current safeguards do not yet guarantee total reliability.

Finally, different scenarios are projected for the year 2030, highlighting the need for ongoing scientific oversight to address technological uncertainty.

Second installment

This report is the second international scientific assessment of capabilities and focuses on emerging risks of general-purpose AI. Its goal is to help policymakers address the “evidence dilemma”: the need to make decisions about a rapidly evolving technology, even when data on its impacts are slow to emerge.

Here are three key takeaways from this edition:

• It analyzes the evolution of vision models and autonomous agents.

• An in-depth analysis of loss of control and cybersecurity in cutting-edge models.

• The need for clear safety thresholds before deploying dangerous capabilities.