Inside Cisco’s 2026 AI Security Report: MCP exploits, Agent attacks & new threats

Olena's Data & Engineering Corner · Beginner ·🤖 AI Agents & Automation ·2mo ago

Skills: Agent Foundations90%Tool Use & Function Calling80%Multi-Agent Systems70%Autonomous Workflows60%

0:00 intro 0:41 Most interesting recent exploits 01:54 Prompt injections 03:08 AI supply chain 04:38 Agent skills 05:42 Claude code used for cyber-espionage 06:57 MCP 09:18 State-sponsored threat actors 10:24 Most concerning scenarios 10:24 The new paradigm and AI security framework 15:01 Policies 15:41 Tooling: Agentic Skills Scanner 16:29 Multi-turn jailbreak attacks 18:03 MCP and A2A scanners 19:08 SecureBERT 2.0 https://www.cisco.com/site/us/en/products/security/state-of-ai-security.html

Watch on YouTube ↗ (saves to browser)

Sign in to unlock AI tutor explanation · ⚡30

More on: Agent Foundations

View skill →

Build and Deploy an Agent with Reasoning Engine in Vertex AI

Adding a Phone Gateway to a Virtual Agent

From Zero to Working AI Agent in 60 Seconds

From Zero to Working AI Agent in 60 Seconds

Create An AI Agent With Replit That Automates Your Sales

Create An AI Agent With Replit That Automates Your Sales

Capstone: Autonomous Runway Detection for IoT

Capstone: Autonomous Runway Detection for IoT

AI Agents with Model Context Protocol & Typescript

AI Agents with Model Context Protocol & Typescript

Related AI Lessons

Why AI builders need infrastructure that actually scales

AI builders need scalable infrastructure to avoid production failures, learn how to bridge the architecture gap

WAB web agent bridge

Learn how WAB web agent bridge addresses trust and security for AI agents in e-commerce transactions

Human-in-the-Loop: The Runtime Enforcement of requires_approval

Learn how to implement a runtime enforcement mechanism, requires_approval, to prevent AI agents from taking unwanted actions

Negotiating With a Toaster That Wants to Be a Spreadsheet

Learn how to approach unusual AI interactions and improve your understanding of AI behavior

Chapters (15)

intro

0:41 Most interesting recent exploits

1:54 Prompt injections

3:08 AI supply chain

4:38 Agent skills

5:42 Claude code used for cyber-espionage

6:57 MCP

9:18 State-sponsored threat actors

10:24 Most concerning scenarios

10:24 The new paradigm and AI security framework

15:01 Policies

15:41 Tooling: Agentic Skills Scanner

16:29 Multi-turn jailbreak attacks

18:03 MCP and A2A scanners

19:08 SecureBERT 2.0

I Built an AI Tool That Tells You Why Your Resume Gets Rejected #python #agenticai #groq

ChethanAIChronicles