Skip to content
i-ise.com·Indexed by this site · Originally publishedDate not given· Governance →· Its place in the scan record →SelectedAttention42

Institute of International Socio-Economic Research report: two major topics, frontier model access safeguards and loss-of-control risk

IISE研究员讨论前沿模型获取与失控风险治理

AI summary

In a November 2026 report, Koizumi Yusuke, principal researcher at the Institute of International Socio-Economic Research, reviewed international developments over roughly the past six months on two major topics: frontier model access safeguards and loss-of-control risk. On access, Anthropic in April released Mythos Preview, which has outstanding vulnerability discovery and exploitation capabilities, on a limited basis through Project Glasswing, and about 150 organizations in Japan and more than 15 countries gained access in June; the United States signed EO 14409 on June 2, which provides for pre-release evaluation of targeted frontier models for up to 30 days through a voluntary framework, and had previously required Anthropic to suspend foreign nationals' access to Fable 5 and Mythos 5 on export-control grounds. The report recommends that Japan amend its AI Promotion Act with reference to the EU AI Act, introducing enforcement tools such as information-provision requirements and fines, and coordinate with the EU to maintain negotiating leverage with the United States.

Why it was selected

From a Japanese think tank perspective, the report reviews governance developments in the contest over frontier model access and in loss-of-control incidents, and proposes amending the law to strengthen regulation and coordinating with the EU to enhance negotiating leverage with the United States.

Read originali-ise.com