Second key update to the AI Safety Report: Technical and organizational approaches to risk management

The field of AI is developing dynamically. In light of this constant change, the committee behind the International Scientific Report on the Safety of Advanced AI has introduced key updates to address the safety of AI. These updates will allow the latest AI developments to be addressed promptly between comprehensive annual reports. The second interim report deals with the technical precautions and protective measures for general-purpose AI developed by researchers and companies. It also outlines remaining vulnerabilities. Prof. Antonio Krüger, CEO of the German Research Center for Artificial Intelligence, contributed to the report in his capacity as the German representative appointed by the German government to the Expert Advisory Panel of the AI Safety Report.

In 2023, the British government commissioned the International Scientific Report on the Safety of Advanced AI. In January 2025, ahead of the AI Action Summit in Paris, the final report was published. Given the constant evolution of AI, the international team of researchers led by Yoshua Bengio has now published its second interim report.

Since the first AI Safety Report in 2025, the debate on AI safety has changed. More companies have established their own safety measures, known as "Frontier AI Safety Frameworks”. Governments and international organizations have created frameworks for regulating general-purpose AI. Researchers and developers have been able to refine techniques for training safe models and detecting AI-generated content. However, attackers can still circumvent defensive measures, and the effectiveness of many safety precautions has not yet been thoroughly researched.

More details can be found in the second Key Update of the AI Safety Report 2025.

Second key update to the AI Safety Report: Technical and organizational approaches to risk management

Press contact:

Jennifer Oberhofer

Valentina Röther, M.A.