У нас вы можете посмотреть бесплатно Dreadnode at Offensive AI Con (OAIC) 2025 | From Benchmarks to Breaches: Scaling Offensive Security или скачать в максимальном доступном качестве, видео которое было загружено на ютуб. Для загрузки выберите вариант из формы ниже:
Если кнопки скачивания не
загрузились
НАЖМИТЕ ЗДЕСЬ или обновите страницу
Если возникают проблемы со скачиванием видео, пожалуйста напишите в поддержку по адресу внизу
страницы.
Спасибо за использование сервиса ClipSaver.ru
As offensive AI agents move from demos to production, they face a critical bottleneck: the very abstractions that make them capable limit their performance as models improve. At the inaugural Offensive AI Con Dreadnode's Nick Landers (CTO & Co-Founder) and Shane Caldwell (Principal Research Engineer) presented a comprehensive approach to scaling offensive AI security through test-time verification, domain-specific training, and synthetic data generation. Presentation slides here: https://github.com/Offensive-AI-Con/O... In this session, explore how verification provides a foundation for LLM performance on complex tasks, from competitive programming to penetration testing. Learn how programmatic verifiers and judge models enable autonomous success signals, dramatically reducing false positives while boosting performance over long horizons. This talk covers: → Test-time verification fundamentals and how they work in practice → Real-world demonstrations across credential finding, privilege escalation, and lateral movement → A complete taxonomy of verifiers: environment-based, regex-based, and LLM-judge approaches → Training private models that match frontier performance using verified samples and GRPO → Internal benchmarks comparing closed frontier models to fine-tuned local models +++++ About Dreadnode: Dreadnode is an offensive AI company, redefining what’s possible when AI is applied to offense at scale. We've built the most advanced platform for offensive cyber operations—giving AI red teams, security engineers, and model developers the tools to build, evaluate, and deploy agents and develop capabilities with complete transparency and confidence. Backed by Decibel, In-Q-Tel, NFC, Sands Capital, Indie VC, Aviso Ventures, Jon Oringer, and leading government, enterprise, and AI lab partners, Dreadnode is advancing the state of offensive security. Learn more at https://dreadnode.io/ or follow us @dreadnode. #OffensiveAI #AIRedTeam #MachineLearning #Cybersecurity #LLM #AIAgents #PenetrationTesting