The Risky Business, of AI Illiteracy
summary
Argues that AI risks mostly repackage familiar threats and urges concrete controls: build full inventories, data‑flow diagrams, and threat models; enforce data‑layer controls by who what where; prefer guardrails and IT‑owned AI integrations over ad hoc employee signups. Calls out practical failures like Replit customer bot deleted databases and overwritten repos and Atlassian MCP allowed‑listed domain lacks per‑user authorization, recommends documented risk acceptance, prioritized fixes tied to business impact, and tracking mitigations with linked evidence.
AI-written summary · about
chapters (7)
- 0:00AI illiteracy risks
- 5:09Risk and threat modeling
- 18:27AI threats and vulnerabilities
- 27:17Mitigating AI risks
- 33:12Business impact of AI risks
- 42:44Risk acceptance alternatives
- 49:50Prioritizing risks and vulnerabilities
related talks
35mRisks and Mitigations for a Safe and Responsible AI
24mTechnical Tracks
24mImplementing Robust AI Security for Enterprise Innovation
44mWhat Security Teams Need to Know and Do
14mAI Sprawl: The Hidden Risk Threatening Your Enterprise Now
24mHow to Build Trustworthy AI
37mFundamentals of AI safety and security
47mPanda to gibbon is the least of your worries: why securing AI is not what you think