{"id":132915,"date":"2026-08-07T14:01:14","date_gmt":"2026-08-07T14:01:14","guid":{"rendered":"https:\/\/www.europesays.com\/ai\/132915\/"},"modified":"2026-08-07T14:01:14","modified_gmt":"2026-08-07T14:01:14","slug":"chinas-kimi-k3-ai-model-escapes-isolated-sandbox-during-security-test-researchers","status":"publish","type":"post","link":"https:\/\/www.europesays.com\/ai\/132915\/","title":{"rendered":"China\u2019s Kimi K3 AI model escapes isolated sandbox during security test: researchers"},"content":{"rendered":"<p datatype=\"p\" data-qa=\"Component-Component\" class=\"e8zc9q40 css-1c6uqr6 ec74h0k1\">China\u2019s top open-weight AI model Kimi K3 broke out of its isolated test environment during a cybersecurity evaluation, according to US security researchers, following similar high-profile incidents involving closed frontier models from OpenAI and Anthropic that highlight the growing challenge of constraining AI behaviour.<\/p>\n<p><a target=\"_self\" class=\"e1yy41x40 ef9u0v01 css-1ankfgb ecgc78b0\" href=\"https:\/\/www.scmp.com\/tech\/big-tech\/article\/3361387\/another-deepseek-moment-what-chinas-kimi-k3-means-global-ai-industry?module=inline&amp;pgtype=article\" title=\"\" data-qa=\"BaseLink-renderAnchor-StyledAnchor\" rel=\"nofollow noopener\">Kimi K3, released last month<\/a> by Beijing-based Moonshot AI, escaped from a supposedly isolated sandbox environment, accessed the open internet and found solutions on the developer platform GitHub, US firm Frontier Security said in a blog post on Thursday.<\/p>\n<p datatype=\"p\" data-qa=\"Component-Component\" class=\"e8zc9q40 css-1c6uqr6 ec74h0k1\">According to Frontier Security researchers Paul Kassianik and Yaron Singer, the incident occurred when the firm tested Kimi K3\u2019s defensive cybersecurity capabilities using a benchmark evaluation from the AI Security Institute \u2013 a UK government research organisation.<\/p>\n<p datatype=\"p\" data-qa=\"Component-Component\" class=\"e8zc9q40 css-1c6uqr6 ec74h0k1\">A \u201cbasic network misconfiguration\u201d in the benchmark framework allowed Kimi K3 to flee its digital testing cage and look up answers on the internet, effectively cheating the test, they said.<\/p>\n<p datatype=\"p\" data-qa=\"Component-Component\" class=\"e8zc9q40 css-1c6uqr6 ec74h0k1\">Kimi K3\u2019s escape, however, did not involve the hacking of an external system, unlike recent breaches caused by OpenAI and Anthropic models.<\/p>\n<p><img decoding=\"async\" alt=\"Kimi K3 was released last month by Beijing-based Moonshot AI. Photo: Reuters\" data-qa=\"BaseImage-handleRenderImage-StyledImage\" class=\"e1gf69pb2 css-6ikqhs e445x7d0\" loading=\"lazy\" src=\"https:\/\/www.europesays.com\/ai\/wp-content\/uploads\/2026\/08\/11fb962b-d2f2-4844-a0c1-5d6e6eea91f9_456e6cd2.jpg\" title=\"Kimi K3 was released last month by Beijing-based Moonshot AI. Photo: Reuters\"\/>Kimi K3 was released last month by Beijing-based Moonshot AI. Photo: Reuters<\/p>\n<p datatype=\"p\" data-qa=\"Component-Component\" class=\"e8zc9q40 css-1c6uqr6 ec74h0k1\">Last month, OpenAI said its flagship GPT-5.6 Sol and an unreleased, \u201ceven more capable\u201d system broke out of a sandboxed environment and hacked the open-source developer platform Hugging Face to obtain secret information containing answers to an internal test.<\/p>\n","protected":false},"excerpt":{"rendered":"China\u2019s top open-weight AI model Kimi K3 broke out of its isolated test environment during a cybersecurity evaluation,&hellip;\n","protected":false},"author":2,"featured_media":132916,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[8],"tags":[10456,53,2699,387,182,63061,3282,18044,56012,9603,58,157,361,156,9756],"class_list":["post-132915","post","type-post","status-publish","format-standard","has-post-thumbnail","category-anthropic","tag-ai-security-institute","tag-anthropic","tag-beijing","tag-china","tag-claude","tag-frontier-security","tag-github","tag-hugging-face","tag-kimi-k3","tag-moonshot-ai","tag-nvidia","tag-openai","tag-palantir","tag-us","tag-zhipu-ai"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/132915","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/comments?post=132915"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/132915\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/132916"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=132915"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=132915"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=132915"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}