{"693304":{"#nid":"693304","#data":{"type":"news","title":"Where Are They Now? Team Atlanta Takes AI Cybersecurity Beyond the Competition","body":[{"value":"\u003Cp\u003E\u003Ca href=\u0022https:\/\/team-atlanta.github.io\/\u0022\u003ETeam Atlanta\u003C\/a\u003E spent two years exploring how artificial intelligence (AI) could help identify and fix software security problems as part of the Defense Advanced Research Projects Agency (DARPA) \u003Ca href=\u0022https:\/\/sites.gatech.edu\/winningaixcc\/\u0022\u003EAI Cyber Challenge\u003C\/a\u003E (AIxCC).\u003C\/p\u003E\u003Cp\u003EAfter a grueling, world-class competition, the team \u003Ca href=\u0022https:\/\/www.cc.gatech.edu\/news\/georgia-tech-makes-history-wins-darpa-challenge\u0022\u003Ewon\u003C\/a\u003E the $4 million first-place prize in August 2025.\u0026nbsp;\u003C\/p\u003E\u003Cp\u003EThe team has since disbanded. Some members have moved into the corporate world, while others continue in academia. Despite these varied paths, Team Atlanta\u2019s search for breakthrough AI cybersecurity solutions continues.\u003C\/p\u003E\u003Ch2\u003EFrom AIxCC to Microsoft\u003C\/h2\u003E\u003Cp\u003ETeam Atlanta lead \u003Cstrong\u003ETaesoo\u003C\/strong\u003E \u003Cstrong\u003EKim\u003C\/strong\u003E, principal security researcher \u003Cstrong\u003EDongkwan\u003C\/strong\u003E \u003Cstrong\u003EKim\u003C\/strong\u003E, and three other former team members joined Microsoft Security after the competition, where they continue to work on AI and cybersecurity in the \u003Ca href=\u0022https:\/\/github.com\/microsoft\/forge-lab\u0022 target=\u0022_blank\u0022\u003EFrontier Offensive Research \u0026amp; Generative Exploitation (FORGE) Lab\u003C\/a\u003E.\u003C\/p\u003E\u003Cp\u003EThe move has given the researchers an opportunity to apply concepts from AIxCC to real-world software. It has also exposed them to challenges that are difficult to replicate in a competition or research lab.\u003C\/p\u003E\u003Cp\u003EDongkwan Kim said that team members continue to share what they learned through research papers, reports, software, and benchmarks. At Microsoft, these lessons are tested against real customer code.\u003C\/p\u003E\u003Cp\u003EThis work led to \u003Ca href=\u0022https:\/\/www.microsoft.com\/en-us\/security\/blog\/2026\/05\/12\/defense-at-ai-speed-microsofts-new-multi-model-agentic-security-system-tops-leading-industry-benchmark\/\u0022\u003EMDASH\u003C\/a\u003E, Microsoft\u0027s multi-agent, multi-model system for vulnerability discovery and remediation. The system uses more than 100 AI agents to search for vulnerabilities, validate potential findings, and help develop fixes.\u003C\/p\u003E\u003Cp\u003ERather than asking a single AI model if a piece of code contains a vulnerability, MDASH can examine a problem using multiple agents. The system can also try to create an exploit to determine whether a potential vulnerability is real and reachable.\u003C\/p\u003E\u003Cp\u003EFor Dongkwan Kim and the other former Team Atlanta members at Microsoft, moving from AIxCC to MDASH showed how research can look different when applied to real software.\u003C\/p\u003E\u003Cp\u003E\u201cAIxCC was not an endpoint for Team Atlanta,\u201d he said. \u201cWe continue sharing what we learned through technical reports, papers, open artifacts, and benchmarks for vulnerability discovery and patching.\u201d\u003C\/p\u003E\u003Cp\u003EWorking with customer code introduced new challenges for the Microsoft team. The researchers had to consider security boundaries, existing development processes, how findings would be reviewed, and how vulnerabilities would be prioritized for action.\u003C\/p\u003E\u003Cp\u003EThese challenges created a new feedback loop between research and product development. The team took ideas from AIxCC, test them on real software, learn where they worked and where they fell short, and use those lessons to improve the system.\u003C\/p\u003E\u003Cp\u003EThe experience also showed that AI does not replace the need for cybersecurity experts. Human researchers are still needed to provide context, decide how to break down difficult problems, validate findings, understand attack paths, and determine the potential impact of a vulnerability.\u003C\/p\u003E\u003Cp\u003ETaesoo Kim, a professor in the \u003Ca href=\u0022https:\/\/scp.cc.gatech.edu\/\u0022\u003ESchool of Cybersecurity and Privacy\u003C\/a\u003E and vice president of security research at Microsoft, said the competition continues to influence the Microsoft team\u0027s work.\u003C\/p\u003E\u003Cp\u003E\u201cEvery single activity that we\u2019ve done at Microsoft was influenced by AIxCC,\u201d he said in a \u003Ca href=\u0022https:\/\/commandline.microsoft.com\/taesoo-kim-interview-mdash-security-research\/\u0022\u003E2026 interview with Microsoft\u003C\/a\u003E.\u003C\/p\u003E\u003Cp\u003EThe work has continued to grow. In 2026, Microsoft introduced \u003Ca href=\u0022https:\/\/microsoft.ai\/models\/mai-cyber-1-flash\/\u0022\u003EMAI-Cyber-1-Flash\u003C\/a\u003E, an AI model built specifically for cybersecurity and designed to work with MDASH.\u003C\/p\u003E\u003Cp\u003EThe model is designed to handle many of MDASH\u0027s tasks while allowing larger AI models to focus on harder problems. Microsoft reported that replacing 80% of the models used by MDASH with MAI-Cyber-1-Flash increased its performance on the CyberGym benchmark from 88.4% to 95.95%.\u003C\/p\u003E\u003Cp\u003EThe benchmark includes more than 1,500 examples from 188 software projects. Microsoft also reported that the new setup can achieve similar performance at half the cost of the previous MDASH configuration.\u003C\/p\u003E\u003Cp\u003E\u201cAIxCC has sparked cybersecurity innovation across the globe, from frontier AI companies to small academic labs,\u201d said Taesoo Kim. \u201cIt is an incredibly exciting time to be working in AI-driven security engineering and research, as we see AI fundamentally reshape how we find and fix vulnerabilities at scale.\u201d\u003C\/p\u003E\u003Ch2\u003EContinuing the Research at Georgia Tech\u003C\/h2\u003E\u003Cp\u003EThe AIxCC experience also continues to shape cybersecurity research at Georgia Tech.\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EJiho\u003C\/strong\u003E \u003Cstrong\u003EKim\u003C\/strong\u003E, a Georgia Tech Ph.D. student who was part of Team Atlanta\u0027s research efforts, said one of the biggest lessons he\u2019s brought back to campus was learning how AI and traditional security tools can work together.\u003C\/p\u003E\u003Cp\u003EOne of these tools is fuzzing. Fuzzing tests software with large numbers of inputs to detect unexpected behavior or security problems. Large language models (LLMs), meanwhile, can reason about code and help guide the search.\u003C\/p\u003E\u003Cp\u003EDuring the AIxCC competition, the Team Atlanta combined these ideas in \u003Ca href=\u0022https:\/\/dl.acm.org\/doi\/10.1145\/3832783.3834365\u0022\u003ECtxFuzz\u003C\/a\u003E, a system that uses an LLM to help guide fuzzing toward parts of software that may contain vulnerabilities.\u003C\/p\u003E\u003Cp\u003EThe research was submitted to the \u003Ca href=\u0022https:\/\/conf.researchr.org\/home\/ase-2026\u0022\u003EACM\/IEEE International Conference on Automated Software Engineering\u003C\/a\u003E, where it received the ACM Distinguished Paper Award.\u003C\/p\u003E\u003Cp\u003E\u201cSince AIxCC, we\u2019ve continued building on some of the core findings from our two-year journey with Team Atlanta,\u201d Jiho Kim said.\u003C\/p\u003E\u003Cp\u003E\u201cOne of the biggest lessons for us was understanding how LLMs and traditional fuzzing can work together in a mutually beneficial way, using each to overcome the limitations of the other and ultimately find more and deeper vulnerabilities.\u201d\u003C\/p\u003E\u003Cp\u003EThe researchers also wanted to make the lessons from AIxCC useful to other cybersecurity researchers.\u003C\/p\u003E\u003Cp\u003EThey created an \u003Ca href=\u0022https:\/\/www.cc.gatech.edu\/news\/competition-community-how-team-atlantas-ai-cybersecurity-breakthrough-going-open-source\u0022\u003EOpen-Source Cyber Reasoning System Framework\u003C\/a\u003E (OSS-CRS) and CRSBench, which provides researchers and security professionals with ways to use, test, and build on technology developed through AIxCC.\u003C\/p\u003E\u003Cp\u003E\u201cWe\u2019ve also been focused on making the results of AIxCC useful to the broader security community,\u201d Kim said. \u201cWe created OSS-CRS and CRSBench so that researchers and practitioners can more easily use, reproduce, and build upon the capabilities and lessons from our AIxCC work.\u201d\u003C\/p\u003E\u003Cp\u003EThe team\u0027s research has also moved beyond identifying vulnerabilities. Researchers are studying how well AI models can understand and reason about computer code.\u003C\/p\u003E\u003Cp\u003EJiho Kim and his lab mates in the \u003Ca href=\u0022https:\/\/gts3.org\/\u0022\u003ESystems Software \u0026amp; Security Lab\u003C\/a\u003E developed COIN, a benchmark designed to test how well LLMs can reason about code. The work has been accepted to the \u003Ca href=\u0022https:\/\/2026.emnlp.org\/\u0022\u003E2026 Conference on Empirical Methods in Natural Language Processing\u003C\/a\u003E.\u003C\/p\u003E\u003Cp\u003EThe Georgia Tech researchers are also developing additional benchmarks to study the various types of reasoning required for cybersecurity. They are using what they learn to create new datasets and methods for training AI models focused on cybersecurity.\u003C\/p\u003E\u003Cp\u003E\u201cWe\u2019re using what we learn from these benchmarks to train cybersecurity-focused models through new datasets and training methodologies,\u201d Jiho Kim said.\u003C\/p\u003E\u003Cp\u003E\u003Cem\u003EPhotos by John Popham\/College of Computing\u003C\/em\u003E\u003C\/p\u003E","summary":"","format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003ETeam Atlanta\u2019s AI cybersecurity work has advanced from a competition victory to real-world vulnerability discovery. Former members continue this work through Microsoft products and Georgia Tech research. Microsoft\u2019s MDASH uses more than 100 AI agents for vulnerability discovery, validation, exploitation testing, and remediation; MAI-Cyber-1-Flash raised CyberGym performance from 88.4% to 95.95% while reducing costs. Georgia Tech researchers combine large language models with fuzzing through CtxFuzz, which received an ACM Distinguished Paper Award, and released OSS-CRS, CRSBench, and COIN to advance cybersecurity research.\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"Team Atlanta\u2019s AI cybersecurity work has advanced from a competition victory to real-world vulnerability discovery. Former members continue this work through Microsoft products and Georgia Tech research."}],"uid":"36253","created_gmt":"2026-10-19 17:56:04","changed_gmt":"2026-10-20 13:38:47","author":"John Popham","boilerplate_text":"","field_publication":"","field_article_url":"","location":"Atlanta, GA","dateline":{"date":"2026-10-20T00:00:00-04:00","iso_date":"2026-10-20T00:00:00-04:00","tz":"America\/New_York"},"extras":[],"hg_media":{"680033":{"id":"680033","type":"image","title":"AIxCC-2025-27-web-copy.jpg","body":null,"created":"1776880174","gmt_created":"2026-04-22 17:49:34","changed":"1776880174","gmt_changed":"2026-04-22 17:49:34","alt":"A group of people standing inside of a convention hall. ","file":{"fid":"264266","name":"AIxCC-2025-27-web-copy.jpg","image_path":"\/sites\/default\/files\/2026\/04\/22\/AIxCC-2025-27-web-copy.jpg","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/2026\/04\/22\/AIxCC-2025-27-web-copy.jpg","mime":"image\/jpeg","size":1078593,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/2026\/04\/22\/AIxCC-2025-27-web-copy.jpg?itok=6t8OfTPC"}},"681466":{"id":"681466","type":"image","title":"TeamAtlanta-Reception-15.jpg","body":null,"created":"1792432685","gmt_created":"2026-10-19 17:58:05","changed":"1792432685","gmt_changed":"2026-10-19 17:58:05","alt":"A crowd of people looking at a screen. The screen shows a team photo of Team Atlanta.","file":{"fid":"265874","name":"TeamAtlanta-Reception-15.jpg","image_path":"\/sites\/default\/files\/2026\/10\/19\/TeamAtlanta-Reception-15.jpg","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/2026\/10\/19\/TeamAtlanta-Reception-15.jpg","mime":"image\/jpeg","size":2223190,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/2026\/10\/19\/TeamAtlanta-Reception-15.jpg?itok=Uk3m9De-"}},"677708":{"id":"677708","type":"image","title":"Team Atlanta at DefCon 33","body":null,"created":"1755285635","gmt_created":"2025-08-15 19:20:35","changed":"1755285635","gmt_changed":"2025-08-15 19:20:35","alt":"Team Atlanta stands on the dark DefCon stage during the convention\u0027s closing ceremony. ","file":{"fid":"261641","name":"IMG_9163.jpg","image_path":"\/sites\/default\/files\/2025\/08\/15\/IMG_9163.jpg","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/2025\/08\/15\/IMG_9163.jpg","mime":"image\/jpeg","size":2393783,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/2025\/08\/15\/IMG_9163.jpg?itok=QL3hjg7_"}},"681468":{"id":"681468","type":"image","title":"AIxCC-2025-38.jpg","body":null,"created":"1792432817","gmt_created":"2026-10-19 18:00:17","changed":"1792432817","gmt_changed":"2026-10-19 18:00:17","alt":"A man standing with his hands out looking at the sky. ","file":{"fid":"265876","name":"AIxCC-2025-38.jpg","image_path":"\/sites\/default\/files\/2026\/10\/19\/AIxCC-2025-38.jpg","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/2026\/10\/19\/AIxCC-2025-38.jpg","mime":"image\/jpeg","size":2528658,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/2026\/10\/19\/AIxCC-2025-38.jpg?itok=XTCMKvSN"}},"681469":{"id":"681469","type":"image","title":"AIxCC-2025-42.jpg","body":null,"created":"1792432817","gmt_created":"2026-10-19 18:00:17","changed":"1792432817","gmt_changed":"2026-10-19 18:00:17","alt":"A man looking at the camera smiling. ","file":{"fid":"265877","name":"AIxCC-2025-42.jpg","image_path":"\/sites\/default\/files\/2026\/10\/19\/AIxCC-2025-42.jpg","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/2026\/10\/19\/AIxCC-2025-42.jpg","mime":"image\/jpeg","size":5433378,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/2026\/10\/19\/AIxCC-2025-42.jpg?itok=uc5jvAL_"}},"681470":{"id":"681470","type":"image","title":"AIxCC-2025-50.jpg","body":null,"created":"1792432817","gmt_created":"2026-10-19 18:00:17","changed":"1792432817","gmt_changed":"2026-10-19 18:00:17","alt":"A man looking up at a screen with his hands clasped in front of him.","file":{"fid":"265878","name":"AIxCC-2025-50.jpg","image_path":"\/sites\/default\/files\/2026\/10\/19\/AIxCC-2025-50.jpg","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/2026\/10\/19\/AIxCC-2025-50.jpg","mime":"image\/jpeg","size":2962205,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/2026\/10\/19\/AIxCC-2025-50.jpg?itok=ydXeVf8n"}},"681467":{"id":"681467","type":"image","title":"AIxCC-2025-05.jpg","body":null,"created":"1792432817","gmt_created":"2026-10-19 18:00:17","changed":"1792432817","gmt_changed":"2026-10-19 18:00:17","alt":"A trophy next to a virtual scoreboard. The scoreboard has virtual stats from the AIxCC competition. ","file":{"fid":"265875","name":"AIxCC-2025-05.jpg","image_path":"\/sites\/default\/files\/2026\/10\/19\/AIxCC-2025-05.jpg","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/2026\/10\/19\/AIxCC-2025-05.jpg","mime":"image\/jpeg","size":4251863,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/2026\/10\/19\/AIxCC-2025-05.jpg?itok=tfIiiSsD"}}},"media_ids":["680033","681466","677708","681468","681469","681470","681467"],"groups":[{"id":"47223","name":"College of Computing"},{"id":"1188","name":"Research Horizons"},{"id":"660406","name":"School of Cybersecurity \u0026 Privacy"},{"id":"660367","name":"School of Cybersecurity and Privacy"}],"categories":[{"id":"194606","name":"Artificial Intelligence"},{"id":"153","name":"Computer Science\/Information Technology and Security"},{"id":"135","name":"Research"},{"id":"134","name":"Student and Faculty"},{"id":"8862","name":"Student Research"}],"keywords":[{"id":"192863","name":"go-ai"},{"id":"187915","name":"go-researchnews"}],"core_research_areas":[{"id":"193655","name":"Artificial Intelligence at Georgia Tech"},{"id":"145171","name":"Cybersecurity"},{"id":"39501","name":"People and Technology"}],"news_room_topics":[],"event_categories":[],"invited_audience":[],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[{"value":"\u003Cp\u003EJohn Popham\u003C\/p\u003E\u003Cp\u003ECommunications Officer at the School of Cybersecurity and Privacy\u003C\/p\u003E","format":"limited_html"}],"email":["jpopham3@gatech.edu"],"slides":[],"orientation":[],"userdata":""}}}