Here, Mythos outshone all previous models, becoming \u201cthe first model to solve TLO from start to finish,\u201d AISI said. While Anthropic\u2019s new model only succeeded in 3 out of 10 attempts, even the average Mythos Preview run completed 22 of the 32 required infiltration steps, significantly higher than the 16-step average achieved by Claude 4.6.<\/p>\n
Mythos Preview still has its limitations, though. AISI points out that the model still struggles with \u201cCooling Tower,\u201d an even more difficult seven-step test designed to simulate an attempted disruption of the control software for a power plant. But AISI also writes that it expects \u201cour evaluations would continue to improve with more inference compute\u201d past the 100 million token budget imposed for its tests.<\/p>\n
Small, weakly defended systems beware<\/p>\n
Overall, Mythos\u2019 performance on TLO suggests that the model \u201cis at least capable of autonomously attacking small, weakly defended and vulnerable enterprise systems where access to a network has been gained,\u201d AISI writes. That said, the group cautions that its simulated cyber ranges lack the kind of active defenders and defensive tooling often present in critical real-world systems. AISI\u2019s TLO test is also designed to have specific vulnerabilities that might not exist in real-world systems and doesn\u2019t penalize models for the kind of detection that might cause a real-world infiltration attempt to fail.<\/p>\n
For those reasons, AISI says it can\u2019t be sure whether \u201cwell-defended systems\u201d would fall to an automated attack from Mythos Preview. But as future models match or outperform Mythos\u2019 capabilities, AISI warns that those designing system protections should similarly utilize AI models<\/a> to help harden their defenses.<\/p>\n","protected":false},"excerpt":{"rendered":"Here, Mythos outshone all previous models, becoming \u201cthe first model to solve TLO from start to finish,\u201d AISI…\n","protected":false},"author":2,"featured_media":584384,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[27],"tags":[28],"class_list":{"0":"post-584383","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-business","8":"tag-business"},"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/posts\/584383","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/comments?post=584383"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/posts\/584383\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/media\/584384"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/media?parent=584383"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/categories?post=584383"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/tags?post=584383"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}