Security panel: Can your platform survive AI?
summary
Models include data files and unsafe serializers like pickle that enable arbitrary code execution, and public registries (Hugging Face) and tools (Mythos) make chainable exploits trivial; teams see proof‑of‑concepts on GitHub and exploit‑db and a realistic 24‑hour exploit window after disclosures. Defenses that actually work are runtime mitigations and virtual patching — eBPF/Tetragon shields, live patching of hypervisors, minimal images from Chainguard or Talos, and platform‑level policy and identity controls to stop agentic shadow AI.
AI-written summary · about
chapters (7)
- 0:00Surviving AI in platform engineering
- 1:57Expert panel introduction
- 3:44AI threat landscape
- 12:52AI code generation security
- 19:32Agentic AI and shadow IT risks
- 23:21Securing platforms in AI age
- 26:10Closing remarks
related talks
18mCommit Secure AI: Embedding Security in Code, Pipelines, and Production
37mHow to Break AI Systems (Before Someone Else Does)
01h 24mOffensive Cybersecurity by AI Promises and Pitfalls
36mAgentic AI Under Attack: Live Demos of Exploits Through Autonomy and Trust
29mThe vulnerability apocalypse: AI broke the security playbook
37mFundamentals of AI safety and security
51mPrompt Injection: When Hackers Befriend Your AI
45mHardening AI/ML Systems - The Next Frontier of Cybersecurity