Fireside Chat - AI Security Vulnerabilities
summary
Panel outlines that multiagent autonomous orgs generate 1,000x more network data and vulnerabilities are surfacing at a 5x pace; demonstrated failure modes include an attacker whitelisting an IP to gain persistent access, exfiltrating HR files via Microsoft Copilot (vendor response took 30 days in one incident) and a database wiped in 9 seconds. Practical remedies: compartmentalize agents, run an AI-SOC with unified ownership, dedicated patching and forensics teams, and deploy AI vs AI defenses plus strict access, autonomy and accountability controls.
AI-written summary · about
chapters (11)
- 0:00Introduction and speakers
- 1:50Agentic AI maturity levels
- 3:14Zero trust vs unlimited trust
- 5:18Attacker perspective on AI systems
- 8:24Real-world AI security breach example
- 9:28Identifying AI-driven anomalies
- 11:14Containing AI system breaches
- 12:59Visibility challenges in AI networks
- 14:02Who owns AI security failures
- 16:06Building secure AI from scratch
- 19:36Key takeaways for CISOs
related talks
21mSecurity in the Age of AI & Copilot
14mAI Sprawl: The Hidden Risk Threatening Your Enterprise Now
42mAI for security - friend or foe?
20mSecurity, and AI... so happy together?
36mAgentic AI Under Attack: Live Demos of Exploits Through Autonomy and Trust
31mAgentic AI Threats: How Multi-Agent Systems Get Exploited
45mDeploy First, Secure Never: AI Between Productivity and Security Chaos
34mAccountability in the Age of AI Agents