[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-study-finds-ai-auditing-agent-beats-codex-at-finding-exploits":10,"sections":35},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":30,"feedback":34,"feedback_at":22,"cost_usd":34,"total_tokens":34},8266,"study-finds-ai-auditing-agent-beats-codex-at-finding-exploits","Study Finds AI Auditing Agent Beats Codex at Finding Exploits","A new benchmark shows a specialized two-agent system outperforms general coding agents at discovering and exploiting security flaws in AI agent frameworks.","Researchers built an AI system that finds security holes in other AI systems, and it beats general-purpose coding agents at the job.\n\nThe tool, called AgentXploit, splits the work into two roles. An Analyzer Agent reads a target repository and traces which attacker-controlled inputs reach sensitive operations, flagging candidate attack paths. An Exploiter Agent then turns those paths into working attacks and adjusts them based on runtime feedback. The researchers tested it against 72 reproducible vulnerabilities across 12 open-source AI-agent systems, a set they packaged as AgentXploit-Bench. Across three runs, it hit a 59.3% end-to-end success rate, versus 38.4% for Codex acting as a general-purpose auditor.\n\nThe gap matters because it isolates where general coding agents actually fall short: not necessarily at writing exploit code, but at the slower, structural work of tracing untrusted input through a codebase to find where it becomes dangerous. Even when the researchers matched Codex's token budget to close some of that gap, AgentXploit still won, 59.3% to 46.3%, suggesting the two-role split itself is doing real work, not just extra compute.\n\nThis is the kind of red-teaming that has to arrive before agentic AI tools get more autonomy over files, APIs, and code execution, not after. Whether a benchmark of 72 bugs generalizes to the messier codebases of production agent frameworks is the open question the paper doesn't answer.","[\"ai-security\",\"red-teaming\",\"ai-agents\",\"vulnerability-research\"]","2026-09-28T04:00:00.000Z","2026-09-28T20:26:14.692Z","2026-09-28T20:26:21.062Z","published",null,[],"security",[26,27,28,29],"ai-security","red-teaming","ai-agents","vulnerability-research",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.31318",0,{"sections":36},[37,42,46,51,56,61,66,71,76,81,86,91,96,101],{"name":38,"slug":39,"count":40,"latest_published_at":41},"AI","ai",4900,"2026-09-28T17:44:43.000Z",{"name":43,"slug":24,"count":44,"latest_published_at":45},"Security",766,"2026-09-28T15:35:23.000Z",{"name":47,"slug":48,"count":49,"latest_published_at":50},"Policy","policy",405,"2026-09-28T17:00:51.000Z",{"name":52,"slug":53,"count":54,"latest_published_at":55},"Deals","deals",269,"2026-09-28T17:42:46.000Z",{"name":57,"slug":58,"count":59,"latest_published_at":60},"Hardware","hardware",191,"2026-09-28T15:45:00.000Z",{"name":62,"slug":63,"count":64,"latest_published_at":65},"Science","science",154,"2026-09-28T13:19:18.000Z",{"name":67,"slug":68,"count":69,"latest_published_at":70},"Consumer Tech","consumer-tech",139,"2026-09-28T17:09:47.000Z",{"name":72,"slug":73,"count":74,"latest_published_at":75},"Software","software",91,"2026-09-25T20:55:00.000Z",{"name":77,"slug":78,"count":79,"latest_published_at":80},"Dev Tools","dev-tools",87,"2026-09-28T16:11:42.000Z",{"name":82,"slug":83,"count":84,"latest_published_at":85},"Startups","startups",80,"2026-09-28T17:50:28.000Z",{"name":87,"slug":88,"count":89,"latest_published_at":90},"General","general",49,"2026-09-28T16:44:57.000Z",{"name":92,"slug":93,"count":94,"latest_published_at":95},"Gaming","gaming",48,"2026-09-25T18:35:21.000Z",{"name":97,"slug":98,"count":99,"latest_published_at":100},"Reviews","reviews",31,"2026-09-28T14:31:34.000Z",{"name":102,"slug":103,"count":104,"latest_published_at":105},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]