International AI Safety Report 2025 Released: Key Findings

The International AI Safety Report 2025 has been officially released, offering a comprehensive analysis of the rapid advancements and associated risks of general-purpose AI (GPAI).

This landmark report, chaired by Turing Award-winning computer scientist Yoshua Bengio, represents a collaborative effort by 100 AI experts from 30 countries, the United Nations, the European Union, and the OECD.
Read more on gov.uk.

Key Findings

Rapid Advancement of GPAI

The report highlights the swift progression in GPAI capabilities, noting that models have evolved from struggling with coherent text generation to now excelling in tasks such as programming, image creation, and complex conversations.
Read the full report.

Emerging Risks

Alongside benefits, the report identifies several risks, including:

• Malicious Use: Potential for AI-driven large-scale disinformation, fraud, and cyber-attacks.
  Source.
• Bias and Discrimination: Concerns about AI systems perpetuating biases, leading to unfair treatment of certain groups.
  Source.
• Economic Disruption: Possibility of significant labor market shifts and economic inequalities resulting from AI integration.
  Source.

Risk Management Challenges

The report emphasizes that current risk mitigation techniques are in their infancy and require substantial improvement to effectively address the identified risks.

This report aims to foster constructive, evidence-based discussions on AI safety, serving as a foundational resource for policymakers worldwide. It underscores the importance of understanding GPAI’s capabilities and risks to harness its benefits while ensuring safety and reliability.

For a detailed exploration of the report, you can access the full document here.