AI & ML2026-05-09
MCP Security Risks: Tool Poisoning, Prompt Injection, and the New AI Agent Attack Surface
Model Context Protocol integrations give agents access to tools, files, and services. That power creates new risks: tool poisoning, prompt injection, overbroad permissions, and untrusted server abuse.
10 min read
AI & ML2026-05-04
Security LLM and Agent Workflows: When (and How) to Check Malicious Domains, IPs, and URLs Before Acting
AI assistants in SOAR, IDEs, and browser extensions can exfiltrate data or run malicious code if they fetch the wrong link. This guide gives guardrails: schema for tool calls, policy tiers, and where threat intelligence checks belong in the loop.
5 min read