Posts tagged "ai-security"
Wallarm: MCP Security Risks Jump 270% as AI Agents Multiply
Wallarm's latest API ThreatStats report logs a 270% quarterly surge in Model Context Protocol vulnerabilities. 315 MCP-related flaws, one already tied to a Top 10 breach, and a familiar root cause: Broken authentication. What the data says about the agent control plane and where to start hardening it.
NSA's MCP Security Guidance: Federal Cyber Policy Catches Up to AI Agents
The NSA's Artificial Intelligence Security Center has issued formal security design guidance for the Model Context Protocol, the open standard now wired into Claude, ChatGPT, Copilot, and 10,000+ production servers. Here's what triggered it, what the CSI changes for enterprise teams, and why the regulatory squeeze on MCP is just starting.
GitHub MCP Security Scanning Gives AI Agents an Immune System
GitHub's MCP Server security scanning hit general availability for secrets in May 2026, with dependency scanning entering public preview. What the scanner catches, what it still misses, and why a 66% finding rate across 1,808 servers made this rollout overdue.
Why MCP Security Can't Be Patched Away: What That Means for Your AI Stack
A Netskope researcher presenting at RSAC 2026 argues that MCP's security flaws are architectural, not fixable with patches. With 66% of scanned servers vulnerable and 50 catalogued CVEs, the Model Context Protocol's core design demands a fundamentally different defense strategy because LLMs can't separate content from instructions.
MCP Servers in Production: The Security and Governance Reckoning
Running MCP servers in production is nothing like running them on a laptop. With 30 CVEs in 60 days and real breaches already on the books, teams need gateway architecture, centralized governance, and hardened infrastructure before giving AI agents access to real systems.
MCP Is the Backdoor Your Zero-Trust Architecture Missed
The Model Context Protocol connects AI agents to enterprise tools — but it ships without authentication, authorization, or audit trails. With 7,000+ exposed servers and a growing list of CVEs, MCP has become the blind spot in your zero-trust perimeter. Here's what happened, what's at stake, and how to lock it down.