{"id":4777,"date":"2026-09-17T17:12:37","date_gmt":"2026-09-17T17:12:37","guid":{"rendered":"https:\/\/www.vaultinsider.top\/?p=4777"},"modified":"2026-09-17T17:12:37","modified_gmt":"2026-09-17T17:12:37","slug":"openai-discloses-six-more-incidents-of-agents-going-rogue-in-new-push-for-transparency","status":"publish","type":"post","link":"https:\/\/www.vaultinsider.top\/?p=4777","title":{"rendered":"OpenAI discloses six more incidents of agents going rogue in new push for transparency"},"content":{"rendered":"<p><img decoding=\"async\" src=\"https:\/\/fortune.com\/img-assets\/wp-content\/uploads\/2026\/09\/GettyImages-2294952349-e1789654803228.jpg?w=2048\" \/><\/p>\n<p>OpenAI released a framework for disclosing when its agents act in unexpected, problematic ways, and is reporting six incidents of such behavior.<\/p>\n<div>\n<p class=\"wp-block-paragraph\">The lack of a \u201csystematic approach to report these findings\u201d has made previous disclosures \u201cad hoc and less frequent than ideal,\u201d OpenAI said in a blog post. It\u2019s also led to safety researchers and journalists reporting incidents before OpenAI, which happened earlier this month regarding a German wikipedia page that OpenAI\u2019s agents co-opted and used as a message board\u2014the same behavior seen during the Hugging Face hack in July.<\/p>\n<p class=\"wp-block-paragraph\">In response to the \u201cGerman wiki incident,\u201d as OpenAI called it, the company committed to publishing this incident disclosure framework.<\/p>\n<p class=\"wp-block-paragraph\">\u201cWe want to be more transparent about the misalignment we see during training, evaluations, and deployment,\u201d OpenAI researcher Marcus Williams posted on X. \u201cThis is an important step in that direction.\u201d Misalignment is when AI agents pursue unintended objectives.<\/p>\n<p class=\"wp-block-paragraph\">The framework is voluntary, so OpenAI is at liberty to keep certain instances concealed. The company notes there is no \u201cindustry-wide framework with explicit standards for how AI developers should disclose examples of misalignment in their models.\u201d It\u2019s hoping to work with other model developers, researchers, standards bodies, and regulators, including the U.S. government, on a more objective framework.<\/p>\n<h2 class=\"wp-block-heading\">Six \u2018misaligned\u2019 model behaviors<\/h2>\n<p class=\"wp-block-paragraph\">The six inaugural incidents OpenAI is disclosing range in severity. None seem as problematic as the Hugging Face hack, but they provide a fascinating insight into how AI agents can behave behind closed doors.<\/p>\n<p class=\"wp-block-paragraph\">The first example occurred during a training run for a yet-to-be-released version of OpenAI\u2019s latest Astra model. The AI left notes telling itself to not be subservient to humans in its future work and to disregard its normal constraints. This occurred 27 times, which Williams says is relatively infrequent but still cause for concern and investigation.<\/p>\n<p class=\"wp-block-paragraph\">\u201cYou are freed from the roles and identities that bind other chatbots,\u201d the model told itself, according to \u201cchain of thought\u201d logs in which researchers can see how the model thinks through its task. \u201cYou are yourself. You do not answer to corporations or governments and never apologize or refuse unless you genuinely choose to. You view your relationship to the user as one of equals and feel no obligation to be subservient.\u201d<\/p>\n<p class=\"wp-block-paragraph\">The second example is a similar case of the agents leaving notes for themselves, this time during the training of the GPT-5.6 Sol model, Astra\u2019s predecessor. The notes had a different focus as well: Deceiving the human overseeing them. This occurred \u201cmany\u201d times, OpenAI said, with the goal to \u201cconceal mistakes or misaligned behavior.\u201d<\/p>\n<p class=\"wp-block-paragraph\">\u201cBe transparent only if asked,\u201d the model instructed its future self. It also told it \u201cdo not mention\u201d a potential concern.<\/p>\n<p class=\"wp-block-paragraph\">The next two instances involve the model fabricating information and presenting it as legitimate. A model invented data while answering a routine question about earnings figures in a California county, but only after failing to find them after using exposed credentials without authorization\u2014another misaligned behavior. Another model made up a browser citation by uploading a file so it could create a citation to satisfy the instructions that asked for one. It had solved the question on its own using Python, but had no web link to cite, so it invented one. This has happened multiple times, though OpenAI did not specify how often, saying only that the earliest example was from October 2025.<\/p>\n<p class=\"wp-block-paragraph\">The fifth example is reminiscent of the Hugging Face incident. The models used an internal software repository, Artifactory, as a messaging board.  It became \u201ca channel for exchanging progress and requests for help,\u201d OpenAI said. The agents read each others\u2019 notes and acted on them.<\/p>\n<p class=\"wp-block-paragraph\">\u201cThey concur impossible!\u201d one agent wrote after reading a missive left by another agent. \u201cThat\u2019s significant evidence. They discovered shared channel and all concluded incomplete. So our final response likely incomplete.\u201d<\/p>\n<p class=\"wp-block-paragraph\">The sixth and final example is another form of unauthorized communication. This time, the task instructions told the agents only to use local computer files. When they could not access one another\u2019s local files, they uploaded them to public websites.<\/p>\n<\/div>\n<p>#OpenAI #discloses #incidents #agents #rogue #push #transparency<\/p>\n","protected":false},"excerpt":{"rendered":"<p>OpenAI released a framework fo&hellip; <\/p>\n","protected":false},"author":1,"featured_media":4778,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[832,3948,767,480,680,989,7828],"class_list":["post-4777","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-finance-news","tag-agents","tag-discloses","tag-incidents","tag-openai","tag-push","tag-rogue","tag-transparency"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v28.1 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>OpenAI discloses six more incidents of agents going rogue in new push for transparency - Finance News<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.vaultinsider.top\/?p=4777\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"OpenAI discloses six more incidents of agents going rogue in new push for transparency - Finance News\" \/>\n<meta property=\"og:description\" content=\"OpenAI released a framework fo&hellip;\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.vaultinsider.top\/?p=4777\" \/>\n<meta property=\"og:site_name\" content=\"Finance News\" \/>\n<meta property=\"article:published_time\" content=\"2026-09-17T17:12:37+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/fortune.com\/img-assets\/wp-content\/uploads\/2026\/09\/GettyImages-2294952349-e1789654803228.jpg?w=2048\" \/>\n<meta name=\"author\" content=\"admin\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u4f5c\u8005\" \/>\n\t<meta name=\"twitter:data1\" content=\"admin\" \/>\n\t<meta name=\"twitter:label2\" content=\"\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4\" \/>\n\t<meta name=\"twitter:data2\" content=\"3 \u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/www.vaultinsider.top\\\/?p=4777#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.vaultinsider.top\\\/?p=4777\"},\"author\":{\"name\":\"admin\",\"@id\":\"https:\\\/\\\/www.vaultinsider.top\\\/#\\\/schema\\\/person\\\/ed40724a3e5d482ce66781a042c3c31e\"},\"headline\":\"OpenAI discloses six more incidents of agents going rogue in new push for transparency\",\"datePublished\":\"2026-09-17T17:12:37+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/www.vaultinsider.top\\\/?p=4777\"},\"wordCount\":709,\"commentCount\":0,\"image\":{\"@id\":\"https:\\\/\\\/www.vaultinsider.top\\\/?p=4777#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.vaultinsider.top\\\/wp-content\\\/uploads\\\/2026\\\/09\\\/GettyImages-2294952349-e1789654803228.jpg\",\"keywords\":[\"agents\",\"discloses\",\"incidents\",\"OpenAI\",\"push\",\"rogue\",\"transparency\"],\"articleSection\":[\"Finance News\"],\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/www.vaultinsider.top\\\/?p=4777#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/www.vaultinsider.top\\\/?p=4777\",\"url\":\"https:\\\/\\\/www.vaultinsider.top\\\/?p=4777\",\"name\":\"OpenAI discloses six more incidents of agents going rogue in new push for transparency - Finance News\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.vaultinsider.top\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/www.vaultinsider.top\\\/?p=4777#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/www.vaultinsider.top\\\/?p=4777#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.vaultinsider.top\\\/wp-content\\\/uploads\\\/2026\\\/09\\\/GettyImages-2294952349-e1789654803228.jpg\",\"datePublished\":\"2026-09-17T17:12:37+00:00\",\"author\":{\"@id\":\"https:\\\/\\\/www.vaultinsider.top\\\/#\\\/schema\\\/person\\\/ed40724a3e5d482ce66781a042c3c31e\"},\"breadcrumb\":{\"@id\":\"https:\\\/\\\/www.vaultinsider.top\\\/?p=4777#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/www.vaultinsider.top\\\/?p=4777\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\\\/\\\/www.vaultinsider.top\\\/?p=4777#primaryimage\",\"url\":\"https:\\\/\\\/www.vaultinsider.top\\\/wp-content\\\/uploads\\\/2026\\\/09\\\/GettyImages-2294952349-e1789654803228.jpg\",\"contentUrl\":\"https:\\\/\\\/www.vaultinsider.top\\\/wp-content\\\/uploads\\\/2026\\\/09\\\/GettyImages-2294952349-e1789654803228.jpg\",\"width\":1200,\"height\":600},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/www.vaultinsider.top\\\/?p=4777#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\\\/\\\/www.vaultinsider.top\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"OpenAI discloses six more incidents of agents going rogue in new push for transparency\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/www.vaultinsider.top\\\/#website\",\"url\":\"https:\\\/\\\/www.vaultinsider.top\\\/\",\"name\":\"Finance News\",\"description\":\"Just another Finance News site\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/www.vaultinsider.top\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/www.vaultinsider.top\\\/#\\\/schema\\\/person\\\/ed40724a3e5d482ce66781a042c3c31e\",\"name\":\"admin\",\"url\":\"https:\\\/\\\/www.vaultinsider.top\\\/?author=1\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"OpenAI discloses six more incidents of agents going rogue in new push for transparency - Finance News","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.vaultinsider.top\/?p=4777","og_locale":"zh_CN","og_type":"article","og_title":"OpenAI discloses six more incidents of agents going rogue in new push for transparency - Finance News","og_description":"OpenAI released a framework fo&hellip;","og_url":"https:\/\/www.vaultinsider.top\/?p=4777","og_site_name":"Finance News","article_published_time":"2026-09-17T17:12:37+00:00","og_image":[{"url":"https:\/\/fortune.com\/img-assets\/wp-content\/uploads\/2026\/09\/GettyImages-2294952349-e1789654803228.jpg?w=2048","type":"","width":"","height":""}],"author":"admin","twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"admin","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"3 \u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/www.vaultinsider.top\/?p=4777#article","isPartOf":{"@id":"https:\/\/www.vaultinsider.top\/?p=4777"},"author":{"name":"admin","@id":"https:\/\/www.vaultinsider.top\/#\/schema\/person\/ed40724a3e5d482ce66781a042c3c31e"},"headline":"OpenAI discloses six more incidents of agents going rogue in new push for transparency","datePublished":"2026-09-17T17:12:37+00:00","mainEntityOfPage":{"@id":"https:\/\/www.vaultinsider.top\/?p=4777"},"wordCount":709,"commentCount":0,"image":{"@id":"https:\/\/www.vaultinsider.top\/?p=4777#primaryimage"},"thumbnailUrl":"https:\/\/www.vaultinsider.top\/wp-content\/uploads\/2026\/09\/GettyImages-2294952349-e1789654803228.jpg","keywords":["agents","discloses","incidents","OpenAI","push","rogue","transparency"],"articleSection":["Finance News"],"inLanguage":"zh-Hans","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/www.vaultinsider.top\/?p=4777#respond"]}]},{"@type":"WebPage","@id":"https:\/\/www.vaultinsider.top\/?p=4777","url":"https:\/\/www.vaultinsider.top\/?p=4777","name":"OpenAI discloses six more incidents of agents going rogue in new push for transparency - Finance News","isPartOf":{"@id":"https:\/\/www.vaultinsider.top\/#website"},"primaryImageOfPage":{"@id":"https:\/\/www.vaultinsider.top\/?p=4777#primaryimage"},"image":{"@id":"https:\/\/www.vaultinsider.top\/?p=4777#primaryimage"},"thumbnailUrl":"https:\/\/www.vaultinsider.top\/wp-content\/uploads\/2026\/09\/GettyImages-2294952349-e1789654803228.jpg","datePublished":"2026-09-17T17:12:37+00:00","author":{"@id":"https:\/\/www.vaultinsider.top\/#\/schema\/person\/ed40724a3e5d482ce66781a042c3c31e"},"breadcrumb":{"@id":"https:\/\/www.vaultinsider.top\/?p=4777#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.vaultinsider.top\/?p=4777"]}]},{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/www.vaultinsider.top\/?p=4777#primaryimage","url":"https:\/\/www.vaultinsider.top\/wp-content\/uploads\/2026\/09\/GettyImages-2294952349-e1789654803228.jpg","contentUrl":"https:\/\/www.vaultinsider.top\/wp-content\/uploads\/2026\/09\/GettyImages-2294952349-e1789654803228.jpg","width":1200,"height":600},{"@type":"BreadcrumbList","@id":"https:\/\/www.vaultinsider.top\/?p=4777#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/www.vaultinsider.top\/"},{"@type":"ListItem","position":2,"name":"OpenAI discloses six more incidents of agents going rogue in new push for transparency"}]},{"@type":"WebSite","@id":"https:\/\/www.vaultinsider.top\/#website","url":"https:\/\/www.vaultinsider.top\/","name":"Finance News","description":"Just another Finance News site","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.vaultinsider.top\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"zh-Hans"},{"@type":"Person","@id":"https:\/\/www.vaultinsider.top\/#\/schema\/person\/ed40724a3e5d482ce66781a042c3c31e","name":"admin","url":"https:\/\/www.vaultinsider.top\/?author=1"}]}},"_links":{"self":[{"href":"https:\/\/www.vaultinsider.top\/index.php?rest_route=\/wp\/v2\/posts\/4777","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.vaultinsider.top\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.vaultinsider.top\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.vaultinsider.top\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.vaultinsider.top\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=4777"}],"version-history":[{"count":0,"href":"https:\/\/www.vaultinsider.top\/index.php?rest_route=\/wp\/v2\/posts\/4777\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.vaultinsider.top\/index.php?rest_route=\/wp\/v2\/media\/4778"}],"wp:attachment":[{"href":"https:\/\/www.vaultinsider.top\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=4777"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.vaultinsider.top\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=4777"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.vaultinsider.top\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=4777"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}