APIs are the best protection for your AI strategy -
summary
MCPs proliferate but existing APIs are the real control point: harden the API layer, validate inputs/outputs, and enforce least-privilege scopes so agents swap for short-lived user tokens only when authorized. Real attacks include dynamic “rug‑pull” MCP descriptions, file‑write Trojans, and a WhatsApp scraping that mapped 3.5 billion numbers by sending ~60,000 requests/hour. Expect prompt injections, shadow AI and supply‑chain risks from ~20,000 public MCPs; mitigate with automated tests, daily scans, monitoring of tokens and prompt injections, and an agent manager for traceability.
AI-written summary · about
chapters (10)
- 0:00Welcome and introduction
- 1:25APIs and AI evolution
- 6:07Do you need MCP servers?
- 8:12APIs as AI protection layer
- 10:19Designing AI-ready APIs
- 17:22Security risks in MCP adoption
- 26:58API security lessons for AI
- 35:09Agent identity and scopes
- 42:18Automate testing and monitoring
- 43:41Call to action and closing
related talks
27mSecuring the AI Revolution - [...]
26mGovernance and Security of APIs and MCPs
50mMCP at Scale: Governing AI Agent APIs When Autonomous Systems Are Your Consume…
14mAI Sprawl: The Hidden Risk Threatening Your Enterprise Now
29mMCP Security: Keep Your AI Agents from Spilling the Tea
15mWhy Security Matters: The Risks of Agentic AI and How to Mitigate Them
30mAttack API Architecture
24mDiscussion between Mehdi Medjaoui and Troy Leliard,API Security.