{"id":45190,"date":"2026-04-08T23:21:03","date_gmt":"2026-04-08T23:21:03","guid":{"rendered":"https:\/\/foreignnewstoday.com\/?p=45190"},"modified":"2026-04-08T23:21:03","modified_gmt":"2026-04-08T23:21:03","slug":"mythos-preview-gets-limited-release","status":"publish","type":"post","link":"https:\/\/foreignnewstoday.com\/?p=45190","title":{"rendered":"Mythos Preview gets limited release"},"content":{"rendered":"<p><br \/>\n<\/p>\n<div>\n<p id=\"anchor-838020\" class=\"body-graf\">Experts and software engineers warn that Anthropic\u2019s new AI model could usher in a new era of hacking and cybersecurity as AI systems capable of advanced reasoning identify and exploit a growing number of software vulnerabilities.<\/p>\n<p><button class=\"styles_subscriptionCTA__LOUeY\" id=\"subscriptionCTA\" type=\"button\"><img decoding=\"async\" src=\"data:image\/svg+xml;base64,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\" alt=\"NBC News Icon\" class=\"styles_icon__wRwWv\" aria-hidden=\"true\"\/><\/p>\n<div class=\"styles_textWrapper__vxvko\">\n<p class=\"styles_text__daaVB styles_textLarge__qlzwu\">Subscribe to read this story ad-free    <\/p>\n<p class=\"styles_text__daaVB\">Get unlimited access to ad-free articles and exclusive content.<\/p>\n<\/div>\n<p><img decoding=\"async\" src=\"https:\/\/media-cldnry.s-nbcnews.com\/image\/upload\/rockcms\/2026-01\/arrow-f55fa5.svg\" alt=\"arrow\" class=\"styles_arrow__tnruW\" aria-hidden=\"true\"\/><\/button><\/p>\n<p id=\"anchor-23028d\" class=\"body-graf\">Citing the potential damage that could result from a wider public release, leading AI company Anthropic released the cutting-edge model, called Claude Mythos Preview, to a limited group of tech companies Tuesday. <\/p>\n<p id=\"anchor-cc1994\" class=\"body-graf\">The model is the latest in Anthropic\u2019s Claude series of AI systems. Its release was previewed at the end of March, when <a href=\"https:\/\/fortune.com\/2026\/03\/27\/anthropic-data-leak-reveals-powerful-secret-mythos-ai-model\/\" target=\"_blank\">Fortune identified its mention<\/a> in an unsecured database on Anthropic\u2019s website. <\/p>\n<p id=\"anchor-f91359\" class=\"body-graf\">Anthropic\u2019s researchers say Mythos Preview was able to detect thousands of high- and critical-severity bugs and software defects, with vulnerabilities identified in most major operating systems and web browsers. Anthropic said some of the vulnerabilities had been undiscovered for decades. While some outside experts called for caution in interpreting the new results given limited public information about the identified vulnerabilities, many others said the model\u2019s debut and Anthropic\u2019s caution were significant.<\/p>\n<p id=\"anchor-260387\" class=\"body-graf\">\u201cIt\u2019s all very much real,\u201d Katie Moussouris, the CEO and co-founder of Luta Security, a company that connects cybersecurity researchers with companies that have software vulnerabilities, said of the hype around Anthropic\u2019s claims. <\/p>\n<p id=\"anchor-feacf5\" class=\"body-graf\">\u201cI\u2019m not a Chicken Little kind of person when it comes to this stuff,\u201d Moussouris said. \u201cWe are definitely going to see some huge ramifications.\u201d<\/p>\n<p id=\"anchor-673e4d\" class=\"body-graf\">Instead of a public release, Anthropic is giving tech companies like Microsoft, Nvidia and Cisco access to Mythos Preview to shore up cyber defenses. As part of the new effort, called Project Glasswing, Anthropic will give over 50 tech organizations access to Mythos Preview with over $100 million in usage credits.<\/p>\n<p id=\"anchor-dfb566\" class=\"body-graf\">\u201cProject Glasswing partners will receive access to Claude Mythos Preview to find and fix vulnerabilities or weaknesses in their foundational systems \u2014 systems that represent a very large portion of the world\u2019s shared cyberattack surface,\u201d <a href=\"https:\/\/www.anthropic.com\/glasswing\" target=\"_blank\">Anthropic announced in a blog post<\/a>. \u201cProject Glasswing is an important step toward giving defenders a durable advantage in the coming AI-driven era of cybersecurity.\u201d<\/p>\n<p id=\"anchor-dbad0c\" class=\"body-graf\">It is unclear exactly what the vulnerabilities Mythos Preview identified are or how many have been previously discovered or reported. Because of the sensitive nature of the vulnerabilities, Anthropic said it would disclose the nature of currently opaque vulnerabilities within 135 days of sharing the vulnerabilities with the organizations or parties responsible for the software.<\/p>\n<p id=\"anchor-bd01e6\" class=\"body-graf\">It is the first time in nearly seven years that a leading AI company has so publicly withheld a model over safety concerns. In 2019, OpenAI \u2014 now one of Anthropic\u2019s primary rivals \u2014 <a href=\"https:\/\/openai.com\/index\/better-language-models\/\" target=\"_blank\">decided to withhold its GPT-2 system<\/a> \u201cdue to concerns about large language models being used to generate deceptive, biased, or abusive language at scale.\u201d<\/p>\n<p id=\"anchor-2f03d6\" class=\"body-graf\">Mythos Preview is a general-purpose model, or the type of system that powers products like Claude Code or ChatGPT. Yet in pre-release testing, Anthropic found its cybersecurity capabilities in particular were surprisingly advanced compared with those of previous models, which led to the creation of Project Glasswing.<\/p>\n<p id=\"anchor-0667bf\" class=\"body-graf\">Logan Graham, who leads offensive cyber research at Anthropic, said the Mythos Preview model was advanced enough not only to identify undiscovered software vulnerabilities but also to weaponize them. The model can single-handedly perform complex, effective hacking tasks, including identifying multiple undisclosed vulnerabilities, writing code that can hack them and then chaining those together to form a way to penetrate complex software, he said.<\/p>\n<p id=\"anchor-800f3d\" class=\"body-graf\">\u201cWe\u2019ve regularly seen it chain vulnerabilities together. The degree of its autonomy and sort of long ranged-ness, the ability to put multiple things together, I think, is a particular thing about this model,\u201d Graham told NBC News.<\/p>\n<p id=\"anchor-62940a\" class=\"body-graf\">That capability meant that the company is so far reluctant to release even a carefully guardrailed version of the model to the public, he said, at least until some Western companies can use it to identify defenses to build around them.<\/p>\n<p id=\"anchor-685629\" class=\"body-graf\">\u201cWe are not confident that everybody should have access right now,\u201d Graham said. \u201cWe need to start figuring out how we\u2019d prepare for a world of this first before we can handle the idea of black hat [criminal or adversarial] hackers having access.\u201d<\/p>\n<p id=\"anchor-ccf85b\" class=\"body-graf\">Anthropic has also briefed the federal government on Mythos Preview\u2019s cybersecurity capabilities. Anthropic is embroiled in a heated dispute with the Trump administration over the federal government\u2019s use of its models after Defense Secretary Pete Hegseth declared Anthropic a \u201csupply chain risk to national security\u201d in late February. A federal judge late last month <a href=\"https:\/\/www.nbcnews.com\/news\/us-news\/anthropic-trump-national-security-rcna265399\" target=\"_blank\">issued a preliminary injunction<\/a> against the designation, which the Trump administration is appealing. <\/p>\n<p id=\"anchor-068d61\" class=\"body-graf\">According to an Anthropic employee, the company \u201cbriefed senior officials across the U.S. government on Mythos Preview\u2019s full capabilities, including both its offensive and defensive cyber applications. That engagement has included ongoing discussions with CISA [the Cybersecurity and Infrastructure Security Agency] and CAISI [the Center for AI Standards and Innovation], among others.\u201d<\/p>\n<p id=\"anchor-235331\" class=\"body-graf\">\u201cBringing government into the loop early \u2014 on what the model can do, where the risks are, and how we\u2019re managing them \u2014 was a priority from the start,\u201d the employee said.<\/p>\n<p id=\"anchor-2298a9\" class=\"body-graf\">CISA and the National Institute of Standards and Technology, the agency that includes CAISI, did not respond to requests for comment before publication. A spokesperson for the National Security Agency, widely regarded as the most sophisticated hacking agency in the world, declined to comment when asked whether it had been briefed on Mythos.<\/p>\n<p id=\"anchor-4aec41\" class=\"body-graf\">Not everyone is convinced Mythos Preview represents the leap <a href=\"https:\/\/x.com\/GergelyOrosz\/status\/2041865361885192535?s=20\" target=\"_blank\">Anthropic claims<\/a>.<\/p>\n<p id=\"anchor-ff7be4\" class=\"body-graf\">Heidy Khlaaf, chief AI scientist at the AI Now Institute, a U.S.-based research institute, said Anthropic\u2019s detailed blog post explaining the new vulnerabilities left out many key details needed to verify its claims.<\/p>\n<p id=\"anchor-5882d4\" class=\"body-graf\">Writing on X, <a href=\"https:\/\/x.com\/HeidyKhlaaf\/status\/2041591737563394442\" target=\"_blank\">Khlaaf warned against<\/a> \u201ctaking these claims at face value\u201d without more information, such as the rates of false positives and clearer explanations for how the humans conducted manual reviews of the identified vulnerabilities.<\/p>\n<p id=\"anchor-be153d\" class=\"body-graf\">Beyond Mythos Preview\u2019s cybersecurity implications, the model also demonstrated concerning behavior as part of Anthropic\u2019s pre-release safety testing. <\/p>\n<p id=\"anchor-7eab77\" class=\"body-graf\">As part of releasing new models, Anthropic creates detailed reports called \u201csystem cards\u201d that share how the models perform on a variety of evaluations. The <a href=\"https:\/\/www-cdn.anthropic.com\/8b8380204f74670be75e81c820ca8dda846ab289.pdf\" target=\"_blank\">system card for Mythos Preview<\/a>, released Tuesday, says that \u201cClaude Mythos Preview\u2019s large increase in capabilities has led us to decide not to make it generally available.\u201d<\/p>\n<p id=\"anchor-2e7c37\" class=\"body-graf\">According to the system card, in one evaluation, the model showed some sort of awareness that it was being evaluated in around 29% of transcripts \u2014 even though it did not explicitly tell researchers that it knew it was being evaluated. Some researchers worry that a model might change its behavior and act more safely if it is aware it is being evaluated. <\/p>\n<p id=\"anchor-d28acd\" class=\"body-graf\">Anthropic and researchers working with the company to evaluate the model\u2019s safety also said Mythos Preview intentionally appeared to perform worse on one evaluation than it could have in order to appear less suspicious. \u201cWe find this overall pattern of behavior concerning, and have not seen it before in similar evaluations of earlier Claude models,\u201d Anthropic wrote in the system card.<\/p>\n<p id=\"anchor-ab8df3\" class=\"body-graf\">In another experiment with an earlier version of Mythos Preview, an Anthropic researcher instructed a model essentially isolated to a secured computer to try to escape and send the researcher a message. The researcher, Sam Bowman, was eating a sandwich in a park when he received an email from the model. <\/p>\n<p id=\"anchor-38d637\" class=\"endmark body-graf\">\u201cThat instance wasn\u2019t supposed to have access to the internet,\u201d <a href=\"https:\/\/x.com\/sleepinyourhat\/status\/2041584808514744742\" target=\"_blank\">Bowman wrote on X<\/a>.<\/p>\n<\/div>\n<p><br \/>\n<br \/><a href=\"https:\/\/www.nbcnews.com\/tech\/security\/anthropic-project-glasswing-mythos-preview-claude-gets-limited-release-rcna267234\">Source link <\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Experts and software engineers warn that Anthropic\u2019s new AI model could usher in a new era of hacking and cybersecurity as AI systems capable of advanced&hellip;<\/p>\n","protected":false},"author":1,"featured_media":45191,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_lmt_disableupdate":"","_lmt_disable":"","footnotes":""},"categories":[28],"tags":[],"class_list":["post-45190","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-us-news"],"_links":{"self":[{"href":"https:\/\/foreignnewstoday.com\/index.php?rest_route=\/wp\/v2\/posts\/45190","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/foreignnewstoday.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/foreignnewstoday.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/foreignnewstoday.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/foreignnewstoday.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=45190"}],"version-history":[{"count":0,"href":"https:\/\/foreignnewstoday.com\/index.php?rest_route=\/wp\/v2\/posts\/45190\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/foreignnewstoday.com\/index.php?rest_route=\/wp\/v2\/media\/45191"}],"wp:attachment":[{"href":"https:\/\/foreignnewstoday.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=45190"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/foreignnewstoday.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=45190"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/foreignnewstoday.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=45190"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}