Technical reports · 2024.05

International Scientific Report on the Safety of Advanced AI: Interim Report

International Scientific Report on the Safety of Advanced AI: Interim Report

Overview

This May 2024 interim report reviews assessment and mitigation of general-purpose AI risks, covering benchmarks and adversarial evaluations, safety engineering, robustness, and deployment monitoring and intervention, alongside the limitations of then-available methods.

Publication and version information

DSIT 2024/009, May 2024. The English report has 132 physical pages; the official Chinese link provides only a seven-page executive summary. Citation authors follow the report’s 76-name recommended citation. Yoshua Bengio chaired the report, Sören Mindermann was scientific lead, and the 16-person writing group was a separate role. The Chinese summary truncates the watermarking discussion on p4 and mistranslates biological-risk uncertainty on p5; consult the English original for technical interpretation.

Author-supplied source information

Author-supplied information on arXiv.

DSIT 2024/009; 132 physical English pages. Authors follow the 76-name recommended citation on physical p131. The 16-member writing group is a distinct contributor role. Official Chinese executive summary: 7 physical pages, with documented translation defects.

Version: Interim report (May 2024)

PARTICIPATION ENQUIRY

Participation enquiry

Your enquiry will be delivered by email. We will reply to the address you provide.

Email enquiries: info@battermachine.com