The vulnerability apocalypse: AI broke the security playbook
summary
AI now finds chains of exploits and fuzzes at orders-of-magnitude speed, outpacing human remediation and moving the bottleneck from discovery to fixing; Project Glasswing produces zero validated CVEs despite source access and researchers report models like Opus 47 and GP55 matching automated finders. The actionable path is platform work: build inventories, SBOMs and call graphs from Git history, declarative infra for rapid recovery, stronger segregation, detection, and AI-assisted remediation pipelines. Project Glasswing yielded zero validated CVEs SBOMs and call graphs from Git history agentic insider threats and model poisoning
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:20Panelist intros
- 5:39Vulnerability remediation problem
- 9:01AI and platform engineering
- 12:10Myth of CVEs and exploit chains
- 15:47AI scale imbalance
- 18:51Platform engineers' key role
- 21:43Trust in first-party software
- 23:14AI for secure code?
- 25:33Next bottleneck
- 27:06Poisoned AI agents risk
- 28:55Q&A
related talks
39mThe AI Arms Race in Vulnerability Management, Who’s Winning?
20mThe AI bugpocalypse is here. Now what?
20mSecurity, and AI... so happy together?
43mThe State of AppSec in the Era of AI [Panel]
28mAdversarial Intelligence: Redefining Application Security Through the Eyes of an Attacker
26mSmarter Code, Sneakier Risks: Supply Chain Security in the Age of AI
53mAI-equipped Threat Actors Versus AI-enhanced Cyber Tools: Who Wins?
13mThe Automation of Exploits: Will Eatherton on AI-Driven Security and the End of Obscurity