AI Security
Should an MCP server show different tools to different users?
Yes, if the difference follows per request authorization. The MCP spec lets tool lists vary by granted scopes but not by connection. Why that is safer.
Oct 18, 20263 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
Security for AI or AI for security: what is a vendor actually selling?
Security for AI protects your models and agents; AI for security uses models to help analysts. Learn to tell them apart and ask the right buyer questions.
Oct 18, 20264 min read
AI Security
What should secure coding training cover when developers use AI?
For developers using AI assistants, add review skills, dependency checks, secrets hygiene and policy to training. A syllabus with practice tasks.
Oct 18, 20263 min read
AI Security
Will your SCA tool catch risky packages that AI assistants suggest?
AI coding assistants can suggest unfamiliar or nonexistent packages. What to ask and test when an SCA tool must cover AI-assisted development.
Oct 18, 20263 min read
AI Security
Choosing a SAST tool when AI writes much of your code
AI assistants change code volume and review habits. What to test in a static analysis evaluation: large diffs, pull request fit and secrets coverage.
Oct 18, 20264 min read
AI Security
What to ask a tool about AI assistants adding dependencies
AI assistants add packages fast, sometimes wrong ones. Ask a scanner vendor about checks at suggestion time, in the pull request and after merge.
Oct 18, 20263 min read
AI Security
Prompt rules vs deterministic controls for AI agents
A prompt that says never do X is a request, not a control. Which agent limits belong in code and policy, and which can safely stay in the prompt.
Oct 18, 20263 min read
AI Security
How can a policy reduce prompt injection risk in coding agents?
Policy cannot filter prompt injection away, so it limits what a hijacked agent can do: least privilege, approval for risky actions, untrusted input rules.
Oct 18, 20263 min read
AI Security
Pinning an AI API to a region: a checklist for EU personal data
Sending prompts with personal data to an AI API raises residency and transfer questions. A ten point checklist of what to confirm before you ship.
Oct 18, 20263 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.