AI Security
What log data to redact before sending it to an LLM
Logs hold credentials, personal data and internal names. Here is what to strip or tokenize before an LLM sees them, and how to keep analysis useful.
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
What is prompt injection? Direct vs indirect, and defenses
Prompt injection is input that changes an LLM's behavior in unintended ways. Learn direct vs indirect forms and the OWASP mitigation steps that help.
Oct 18, 20263 min read
AI Security
What is the Model Context Protocol (MCP) and why secure it?
MCP is an open standard for connecting AI applications to data, tools and workflows. Learn how it works and the security questions it raises.
Oct 18, 20263 min read
AI Security
What is MITRE ATLAS and how do AI security teams use it?
MITRE ATLAS is a public knowledge base of adversary tactics and techniques against AI systems. Here is what it contains and how teams put it to work.
Oct 18, 20264 min read
AI Security
What is AIDR? AI detection and response explained
AIDR is an emerging category that watches prompts, agents, models and data pipelines for attacks and misuse. Here is what it covers and what it does not.
Oct 18, 20263 min read
AI Security
What goes in an AI impact assessment: a practical template
An AI impact assessment documents purpose, data, impacts, reassessment, costs, review and risk acceptance. A template from public federal guidance.
Oct 18, 20263 min read
AI Security
A security checklist for MCP servers in VS Code
VS Code can run MCP servers from workspace files, and local servers can run arbitrary code. A checklist on trust, config location, secrets and sandboxing.
Oct 18, 20263 min read
AI Security
Voice cloning and deepfake calls: a verification policy for teams
The FBI warns criminals use AI voice and video to impersonate executives. Set a verification policy so a convincing call cannot move money or access.
Oct 18, 20263 min read
AI Security
USENIX package hallucination study: what the numbers mean
A USENIX Security 2025 paper tested 16 code models and 576,000 samples for invented packages. What the figures say, what they do not, and what to do.
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.