AI Security
Who should own a prompt in production?
Every production prompt needs one accountable owner, a reviewer and a fallback contact. A simple ownership model, a register entry and owner duties.
Oct 19, 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
Who decides what an AI agent may do: the model or the system?
The system must decide, never the model. What complete mediation means for AI agents, where to enforce it, and why prompt rules are not permissions.
Oct 19, 20263 min read
AI Security
Who can add a custom connector on Claude Team and Enterprise?
On Claude Team only Owners add custom connectors, and Enterprise adds custom roles with library access. See who decides, who connects, and what to approve.
Oct 19, 20267 min read
AI Security
Chat assistant, coding agent or scanner: which tool for which job?
A decision guide with twelve common jobs, from explaining an error to finding known flaws in libraries, showing which tool type fits and why.
Oct 19, 20268 min read
AI Security
Which AI coding agent actions should always need human approval?
Require human approval for coding agent actions that are hard to undo, touch secrets or production, change permissions, or reach new network destinations.
Oct 19, 20264 min read
AI Security
Which Claude Code isolation option fits: sandbox, container, VM, or cloud?
Anthropic documents six isolation approaches for Claude Code. A decision guide for choosing between the Bash sandbox, container, VM, and cloud session.
Oct 19, 20266 min read
AI Security
Which AI assistant agent feature should you turn on first?
A rollout order for AI assistant agent features: read-only scheduled tasks, then file generation, desktop apps and remote sandboxes, with exit criteria.
Oct 19, 20266 min read
AI Security
Where does the risk live in a model repo: weights, code or config?
A model repo holds weights, code, tokenizers and config. Each carries a different risk. Here is a map of what to check in each file type before you load.
Oct 19, 20263 min read
AI Security
Where did your few-shot prompt examples come from? A provenance check
Few-shot examples in a prompt teach the model by demonstration. Check where they came from, what they contain and who can change them before you ship.
Oct 19, 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.