AI Is Rewriting the Rules of Cybersecurity
An intimate roundtable on AI-driven pentesting, autonomous vulnerability discovery, and what defenders need to know now.
This event is held under Chatham House Rules. Attendance is by invitation only.
The Landscape Has Shifted
In April 2026, Anthropic unveiled Project Glasswing — a landmark initiative deploying their unreleased frontier model, Claude Mythos, to find zero-day vulnerabilities in the world's most critical open-source software. The results were staggering.
Mythos discovered thousands of previously unknown vulnerabilities: a 27-year-old flaw in OpenBSD, a 16-year-old bug in FFmpeg that automated testing had hit five million times without catching, and critical privilege escalation paths in the Linux kernel. It found exploitable flaws in every major operating system and every major web browser — entirely autonomously, without any human steering.
The window between vulnerability discovery and exploitation has collapsed from months to minutes. This isn't a future scenario — it's happening right now.
What We're Bringing to the Table
Ingram Technologies' cybersecurity research division has been actively investigating the intersection of AI and offensive security. We will present several key findings.
AI-Driven Pentesting: Capabilities & Limits
Our hands-on research into using frontier AI models for penetration testing — what they can already do autonomously, where they still fail, and the narrow window before these capabilities become widely accessible to adversaries.
The Security Cost of AI-Generated Code
AI coding assistants are accelerating development but also introducing new vulnerability patterns at scale. We present data on how AI-written code differs in its security profile from human-written code, and what organizations should do about it.
Glasswing, Mythos & the New Threat Model
What Anthropic's Project Glasswing reveals about the future of vulnerability discovery. We contextualize their findings and discuss what it means for European organizations that cannot afford to wait for the industry to catch up.
Glasswing by the Numbers
Anthropic's Claude Mythos is a step change in AI-assisted security research
CyberGym vulnerability reproduction score — up from 66.6% for prior frontier models
SWE-bench Verified — autonomous software engineering capability
In compute credits committed by Anthropic to defensive security scanning
Event Details
Format
Exclusive VIP dinner and roundtable discussion. An intimate setting for senior security leaders to exchange perspectives on AI's impact on cybersecurity.
Hosted by
Ingram Technologies — Cybersecurity Research Division
Where
Cybersec Europe 2026, Brussels
Rules of Engagement
This dinner is held under Chatham House Rules. Participants may reference insights shared, but may not attribute statements to any individual or organization.
Who This Is For
CISOs & Security Leaders
Decision-makers responsible for defending organizations against an evolving AI-augmented threat landscape.
CTOs & Engineering Leads
Technical leaders adopting AI coding tools who need to understand the security trade-offs being made.
Security Researchers & Red Teams
Practitioners who want to understand how AI changes offensive and defensive operations in practice.
Seats Are Limited
This is a curated, invitation-only dinner for senior leaders. If you are attending Cybersec Europe and want to join the conversation, request your seat now.