{"id":97962,"date":"2026-06-04T15:42:51","date_gmt":"2026-06-04T12:42:51","guid":{"rendered":"https:\/\/forklog.com\/en\/?p=97962"},"modified":"2026-06-04T15:45:15","modified_gmt":"2026-06-04T12:45:15","slug":"researchers-develop-adaptive-ai-worm","status":"publish","type":"post","link":"https:\/\/forklog.com\/en\/researchers-develop-adaptive-ai-worm\/","title":{"rendered":"Researchers Develop Adaptive AI Worm"},"content":{"rendered":"<p>Researchers from the University of Toronto, Vector Institute, University of Cambridge, and ServiceNow have developed a prototype of an adaptive AI worm. The malware generates a unique attack strategy for each target.<\/p>\n<p>The study <a href=\"https:\/\/arxiv.org\/abs\/2606.03811?ref=metacurity.com\">AI Agents Enable Adaptive Computer Worms<\/a> states that the experiment was conducted in an isolated virtual network. The software prototype spread across a heterogeneous network of Linux, Windows, and IoT devices through classic vulnerabilities in corporate networks.<\/p>\n<p>Compromised machines were used by the worm as a computational base for further &#8220;reasoning&#8221; and spreading. The authors claim that this makes the hypothetical cost of infecting a new machine zero for the attacker.<\/p>\n<p>According to developers at <a href=\"https:\/\/cleverhans.io\/latest-research.html\">CleverHans Lab<\/a>, the malware operated on an open language model locally, rather than through a commercial AI platform. The authors emphasize that they did not implement standard virus functions that complicate detection or removal.<\/p>\n<figure class=\"wp-block-image size-full\"><img decoding=\"async\" src=\"https:\/\/forklog.com\/wp-content\/uploads\/img-985f7a41cb665a9b-6178012609418383.webp\" alt=\"image\" class=\"wp-image-281031\"\/><figcaption class=\"wp-element-caption\">AI worm attack scheme. Source: CleverHans Lab.<\/figcaption><\/figure>\n<p>The worm used a recursive cycle of &#8220;reasoning,&#8221; memory, and tools to tailor its tactics to specific devices. Infected nodes with GPUs could provide computing power for further attacks on less powerful devices in the network.<\/p>\n<p>The AI exploited publicly disclosed but still available vulnerabilities, configuration errors, and recurring classes of weaknesses.<\/p>\n<p>The University of Toronto <a href=\"https:\/\/www.utoronto.ca\/news\/u-t-researchers-demonstrate-ai-worm-could-target-any-online-device\">described<\/a> the work as the first demonstration of publicly available AI models powering a virus that changes strategy as it spreads between devices. The material also states that the research aims to help the cybersecurity community prepare for such a threat.<\/p>\n<p>University professor David Lajay called the publication a &#8220;wake-up call&#8221; for developers of protective mechanisms.<\/p>\n<p>Back in September 2025, the threat analysis team at startup Anthropic <a href=\"https:\/\/forklog.com\/en\/news\/anthropic-unveils-first-ai-driven-cyber-espionage-operation\">discovered<\/a> and thwarted the first-of-its-kind cyber espionage campaign conducted by AI.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Researchers from the University of Toronto, Vector Institute, University of Cambridge, and ServiceNow described a prototype of an adaptive AI worm. The malware generates a unique attack strategy for each target.<\/p>\n","protected":false},"author":1,"featured_media":97963,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"select":"1","news_style_id":"1","cryptorium_level":"","_short_excerpt_text":"Researchers developed an adaptive AI worm that tailors attack strategies for each target.","creation_source":"","_metatest_mainpost_news_update":false,"footnotes":""},"categories":[3],"tags":[438,1301,1111],"class_list":["post-97962","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news-and-analysis","tag-artificial-intelligence","tag-blockchain-vulnerabilities","tag-cybersecurity"],"aioseo_notices":[],"amp_enabled":true,"views":"6","promo_type":"1","layout_type":"1","short_excerpt":"Researchers developed an adaptive AI worm that tailors attack strategies for each target.","is_update":"","_links":{"self":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts\/97962","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/comments?post=97962"}],"version-history":[{"count":1,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts\/97962\/revisions"}],"predecessor-version":[{"id":97964,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts\/97962\/revisions\/97964"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/media\/97963"}],"wp:attachment":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/media?parent=97962"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/categories?post=97962"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/tags?post=97962"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}