AI Security
What is ISO/IEC 27090 and what does it mean for AI security?
ISO/IEC 27090 is guidance on security threats to AI systems. See what it covers, where it stands, and how to prepare. Check ISO for the current edition.
Oct 18, 20265 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 you prepare for ISO/IEC 27090 AI security guidance?
Prepare for ISO/IEC 27090 by inventorying AI systems, mapping threats, testing controls and keeping dated evidence. A practical checklist for engineers.
Oct 18, 20265 min read
AI Security
ISO/IEC 27090 vs OWASP LLM Top 10 vs MITRE ATLAS: how they fit
OWASP lists LLM app risks, MITRE ATLAS catalogs attacker behavior, ISO/IEC 27090 is a paid standard on AI threats. Use each for a different job.
Oct 18, 20266 min read
AI Security
Can you use an LLM to triage vulnerability reports safely?
An LLM can sort and summarize vulnerability reports if you treat report text as untrusted, verify its claims with code and keep a human on decisions.
Oct 18, 20264 min read
AI Security
How do you red team an AI agent before launch?
Red team an agent by testing what it can reach, what untrusted content can make it do and whether guardrails hold. Scope, scenarios, pass criteria, fixes.
Oct 18, 20265 min read
AI Security
How should you design tool permissions for an MCP server?
Design MCP tools as narrow single purpose actions with scoped tokens, validated inputs, rate limits and confirmation on risky calls. A practical approach.
Oct 18, 20264 min read
AI Security
ISO 42001 for VPs of engineering: the work behind it
An AI management system turns into engineering work: model inventory, evaluation records, change control and monitoring. Plan these before an audit asks.
Oct 18, 20264 min read
AI Security
ISO 42001 for startup founders: is it worth doing now?
ISO/IEC 42001 is a management system standard for AI. A founder should adopt it when buyers ask or the EU AI Act applies, and start with an AI inventory.
Oct 18, 20264 min read
AI Security
Human approval gates for risky agent actions: a checklist
Which agent actions need a person to approve them, and how do you stop approvals becoming a rubber stamp? A checklist for engineering teams.
Oct 18, 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.