AI Security
How does a chat app learn that a connector added a new tool?
Chat apps ask for the tool list and may subscribe to change notices, but delivery is best effort. What that means for security teams who approve tools.
Oct 19, 20266 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
How do I use Safeguard from Claude or ChatGPT?
Connect Safeguard to Claude or ChatGPT to ask about projects, vulnerabilities and compliance in chat. Setup steps, sample questions and safe habits.
Oct 19, 20268 min read
AI Security
How do I explain AI coding risk to a non-technical boss?
A five minute plain-language brief on AI coding risk for a manager: three risks, one table, one worked example and the decisions you need from them.
Oct 19, 20268 min read
AI Security
Should you give an LLM reachability evidence when triaging CVEs?
A model given only a CVE list guesses at relevance. Give it call path evidence from analysis and it can write sharper, checkable triage notes.
Oct 19, 20263 min read
AI Security
When should destructive AI agent actions need two approvers?
Require two approvers when an agent action is destructive, irreversible or costly. How to define the trigger list and avoid approval fatigue. Steps inside.
Oct 19, 20264 min read
AI Security
How do you test an unknown model file in a no-network container?
Load an unreviewed model in a throwaway container with no network, no secrets and a read-only mount, so a malicious file has nothing to reach or send.
Oct 19, 20264 min read
AI Security
How does OWASP excessive agency apply to a remediation bot?
OWASP's excessive agency risk maps cleanly onto fix bots: too many functions, permissions and autonomy. A walk through each root cause with fixes.
Oct 19, 20264 min read
AI Security
Does the AI show its evidence? An explainability test for buyers
A security AI output you cannot check is a risk. Use this three step test to see whether a feature shows the evidence, the steps, and the limits behind its answer.
Oct 19, 20263 min read
AI Security
Does SCA cover the dependencies of MCP servers and AI agents?
MCP servers and agent code are software with dependencies, so SCA applies. It does not judge what a tool call does. Here is where each control fits.
Oct 19, 20266 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.