Offensive Cybersecurity by AI Promises and Pitfalls
summary
Claims that AI lets attackers and defenders scale exercises instantly and autonomous ethical-hacking bots run large simulated red teams, with one example using an LLM to pull archived HTML and flag five exploitable login flaws on a nuclear-site domain. Demonstrations show GitHub code tweaked by GPTs to accelerate weapons development and polymorphic malware via API that alternates code and language to evade signatures. A survey of ~100 public GPTs finds only one disclosed privacy terms; plugins are pulled for security; proposals center on sandboxing, black-box-style logging and mapping AI use to Cyber Resilience Act / AI Act rules.
AI-written summary · about
chapters (10)
- 0:00Welcome
- 3:57Panel introduction
- 8:13AI in offensive cybersecurity
- 17:41AI-driven attacks
- 29:08Responsibility in AI cybersecurity
- 49:19Regulating AI risks
- 1:00:00Data residency challenges
- 1:05:06AI in cybercrime investigations
- 1:14:17AI risk management frameworks
- 1:23:51Closing remarks
related talks
53mHow AI is impacting cybersecurity. Does it?
49mGenAI and Cybersecurity: Safeguarding the Digital Age
31mHacking AI Applications
49mThe Intersection of AI and Cybersecurity
25mAttack, Defense & Danger in the Age of AI
39mThe Future of Cybersecurity in an AI-driven Era
18mDefending the unknown using AI and cybersecurity
34mAI in Cybersecurity: Offense and Defense