AI Security
Permission-aware retrieval: keep internal search from leaking repos
An internal search index can expose code a user could never open. How to carry repository permissions into retrieval, per OWASP's vector guidance.
Oct 18, 20264 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
OWASP LLM03 supply chain: the ten mitigations as a checklist
OWASP's LLM03:2025 Supply Chain entry lists ten prevention steps. Here they are as a working checklist for model, dataset and library intake in your team.
Oct 18, 20263 min read
AI Security
OpenAI remote MCP tool in the Responses API: approvals and risks
How the Responses API connects to remote MCP servers: configuration fields, approval requests, allowed tools, the connector deprecation and stated risks.
Oct 18, 20263 min read
AI Security
NYC MyCity chatbot gave wrong legal answers: what The Markup found
In March 2024 The Markup reported New York City's business chatbot gave answers that conflicted with the law. The examples, the response, and the lessons.
Oct 18, 20263 min read
AI Security
How to use the NIST AI RMF Playbook without turning it into a checklist
The AI RMF Playbook offers suggested actions per subcategory. Here is how to pick the few that fit your risk, assign owners, and skip the rest on purpose.
Oct 18, 20263 min read
AI Security
NIST AI RMF explained: Govern, Map, Measure and Manage
The NIST AI RMF has four functions: Govern, Map, Measure and Manage. Here is what each one asks a team to do and how they fit together in practice.
Oct 18, 20263 min read
AI Security
The 12 generative AI risks in NIST AI 600-1, explained
NIST's Generative AI Profile names 12 risks unique to or worsened by generative AI. Here is each one in plain words and what to check first.
Oct 18, 20263 min read
AI Security
Multimodal prompt injection: instructions hidden in images and files
Models that read images, PDFs and metadata can be steered by text a person never sees. What OWASP says about multimodal risk and how to reduce exposure.
Oct 18, 20263 min read
AI Security
Multi-turn attacks and persistent memory in LLM apps
Session poisoning, memory persistence and delayed triggers let an attack unfold across turns. What to watch and how to limit what memory can carry forward.
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.