{"id":7837,"date":"2025-12-02T05:26:00","date_gmt":"2025-12-01T20:26:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=7837"},"modified":"2025-11-29T16:29:43","modified_gmt":"2025-11-29T07:29:43","slug":"international-ai-safety-report-2025-second-key-update-technical-safeguards-and-risk-management","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=7837","title":{"rendered":"International AI Safety Report 2025: Second Key Update: Technical Safeguards and Risk Management\u00a0"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>International AI Safety Report 2025: Second Key Update: Technical Safeguards and Risk Management\u00a0<\/strong>[115.9]<br>2025\u5e74\u306e\u56fd\u969bAI\u5b89\u5168\u30ec\u30dd\u30fc\u30c8\u306e\u7b2c2\u306e\u66f4\u65b0\u306f\u3001\u3053\u306e1\u5e74\u3067\u6c4e\u7528AI\u30ea\u30b9\u30af\u7ba1\u7406\u306e\u65b0\u3057\u3044\u5c55\u958b\u3092\u8a55\u4fa1\u3057\u3066\u3044\u308b\u3002 \u7814\u7a76\u8005\u3001\u516c\u5171\u6a5f\u95a2\u3001AI\u958b\u767a\u8005\u304c\u6c4e\u7528AI\u306e\u30ea\u30b9\u30af\u7ba1\u7406\u306b\u3069\u306e\u3088\u3046\u306b\u30a2\u30d7\u30ed\u30fc\u30c1\u3057\u3066\u3044\u308b\u304b\u3092\u8abf\u3079\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2511.19863v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2511.19863v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Tue, 25 Nov 2025 03:12:56 GMT)<\/li>\n\n\n\n<li>AI Safety Report\u306e\u6700\u65b0\u7248\u3002\u30cf\u30a4\u30e9\u30a4\u30c8\u306f\u975e\u5e38\u306b\u53c2\u8003\u306b\u306a\u308b\u304c\u3001\u300cOpen-weight models lag less than a year behind leading closed-weight models, shifting the risk landscape.\u300d\u3068\u3044\u3046\u8a18\u8f09\u306f\u91cd\u8981\u306b\u601d\u3048\u308b\u3002<\/li>\n\n\n\n<li>\u653b\u6483\u9762\u3067\u300ctests show that sophisticated attackers can still bypass safeguards around half of the time when given 10 attempts.\u300d\u3001\u300cAs few as 250 malicious documents inserted into training data can allow attackers to trigger undesired model behaviours with specific prompts. Some research shows that such data poisoning attacks require relatively few resources to carry out, regardless of model size.\u300d\u306a\u72b6\u6cc1\u3060\u304c\u3001\u300cThe number of AI companies with Frontier AI Safety Frameworks more than doubled in 2025: at least 12 companies now have such frameworks.\u300d\u3068\u3044\u3046\u9032\u307f\u5177\u5408\u3082\u8208\u5473\u6df1\u3044\u3002<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[347],"class_list":["post-7837","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-safety"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7837","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=7837"}],"version-history":[{"count":1,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7837\/revisions"}],"predecessor-version":[{"id":7838,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7837\/revisions\/7838"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=7837"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=7837"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=7837"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}