AI Security
Can I try a security connector on a free AI chat plan?
Claude's help center says custom connectors work on Free with one connector, so a trial is possible. What to test, and what a free plan cannot show you.
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
Should you put secrets in a coding agent's sandbox environment?
Environment variables in an agent sandbox are readable by any command the agent runs. When that is acceptable, and safer options to try first.
Oct 19, 20267 min read
AI Security
What is quantization and why does it matter for self-hosted AI cost?
Quantization stores model weights at lower precision to cut memory, so a model may fit on cheaper hardware, at some risk to accuracy. How to test it.
Oct 19, 20267 min read
AI Security
How do I try an AI coding assistant safely in 30 minutes?
A hands-on 30 minute practice run: a throwaway folder, an AI assistant on ask-first settings, a small program, then a security scan to read the results.
Oct 19, 20268 min read
AI Security
What happens when a scheduled AI task fails? Alerts, retries, logs
Scheduled AI tasks can fail without anyone noticing. Ask vendors how failures are alerted, whether retries are safe to repeat, and what the run log shows.
Oct 19, 20266 min read
AI Security
A scheduled AI task misbehaved: what do you do in the first hour?
Pause the task, cut its access, preserve the run history, find what it touched and decide who to tell. A first hour checklist for scheduled AI incidents.
Oct 19, 20266 min read
AI Security
Does a container isolate a coding agent from the host kernel?
Containers separate processes with namespaces but share the host kernel. Here is what that means for a coding agent, and what gVisor and microVMs change.
Oct 19, 20267 min read
AI Security
What should you ask a vendor about their cloud coding agent sandbox?
Twelve questions for any vendor running coding agents in a cloud sandbox: isolation, egress, secrets, git access, retention, and evidence you can request.
Oct 19, 20266 min read
AI Security
How do I check an AI assistant's claim that the fix works?
When an AI says it fixed a bug or vulnerability, trust but verify. Six plain steps: read the change, run tests, rescan, check the version and compare.
Oct 19, 20268 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.