AI Security
Which AI observability metrics matter for security?
Beyond latency and cost, security teams should track tool calls, guardrail blocks, permission denials, retrieval sources and model versions. A starter set.
Oct 18, 20264 min read
Resources
Supply Chain Security, in plain English.
Deep dives, practical guides, and incident analyses from engineers who build Safeguard. No fluff, no vendor FUD — just what you need to ship secure software.
Filter
Articles
RSS feedAI Security
What are the identity risks when an AI coding agent runs as you?
A coding agent that uses your login inherits every permission you have. See the risks, why logs go blind, and how to give agents their own scoped identity.
Oct 18, 20264 min read
AI Security
Is vibe-coded software safe to ship? A security checklist
People vibe code to get working software fast, but speed shifts the review burden. Here is a checklist drawn from NIST SSDF, CISA and OWASP guidance.
Oct 18, 20264 min read
AI Security
What is the OWASP AI Exchange and how should you use it?
The OWASP AI Exchange is an open guide to AI security threats and controls. Use it to threat model AI systems and to pick controls by lifecycle stage.
Oct 18, 20264 min read
AI Security
What is the AI agent attack surface and how do you shrink it?
An AI agent's attack surface is its inputs, tools, memory and credentials. Map each one, cut permissions to the task, and log every tool call.
Oct 18, 20264 min read
AI Security
What is MLSecOps and how does it differ from DevSecOps?
MLSecOps adds security checks to the machine learning lifecycle: data, training, model files and serving. See NIST's attack types and a practical start.
Oct 18, 20264 min read
AI Security
What is MITRE ATLAS and how do you use its matrix for AI systems?
MITRE ATLAS is a public knowledge base of adversary tactics and techniques against AI systems, modeled on ATT&CK. How to read the matrix and apply it.
Oct 18, 20265 min read
AI Security
What is dark AI and how do attackers use AI today?
Dark AI is a loose term for AI used or built for harm. Government assessments say it mostly speeds up known attacks. Here is what that means for defenders.
Oct 18, 20264 min read
AI Security
What is an AI agent gateway and why put one in front of tools?
An agent gateway is a control point between AI agents and their tools. It handles identity, permissions, limits and logs. Here is how to design one.
Oct 18, 20264 min read
Stay informed
Weekly insights on software supply chain security, delivered to your inbox.
Self-healing security runs on Safeguard.
Your first fix PR is minutes away.
No sales call required, even your agent can complete the purchase over MCP.