AI Security
How do you apply least privilege to AI agents?
Give each AI agent its own identity, only the tools it needs, short lived credentials and human approval for risky actions. A checklist for engineers.
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
Are your AI agents secure? Ten attack paths to review
Ten ways AI agents get misused, taken from the OWASP agentic list, with a review question for each. Use it to find which paths your agents leave open.
Oct 18, 20264 min read
AI Security
What is agentic endpoint security and what should you control?
AI agents now run on laptops and servers with real permissions. See what agentic endpoint security covers, the main risks and the controls to set first.
Oct 18, 20264 min read
AI Security
What are agentic browsers and what security risks do they add?
An agentic browser lets an AI model read pages and act for you. The core risk is prompt injection from web content, so limit access and require approval.
Oct 18, 20265 min read
AI Security
What is the difference between agentic AI and generative AI?
Generative AI produces content when asked. Agentic AI plans and acts through tools. That moves the risk from bad text to bad actions. Know the controls.
Oct 18, 20265 min read
AI Security
What should agentic AI security solutions actually do?
Agents act, not just answer, so security shifts to identity, tool access and approvals. Here are the controls to ask any agentic AI security product about.
Oct 18, 20264 min read
AI Security
What is the Agent2Agent (A2A) protocol and how is it secured?
A2A is an open standard for AI agents from different vendors to find each other and share tasks. Learn Agent Cards, auth rules and what to verify.
Oct 18, 20265 min read
AI Security
Why should AI agents use short-lived credentials, and how?
Long-lived keys outlast the task and survive any leak. Give each AI agent its own identity and tokens that expire in minutes, scoped to a single job.
Oct 18, 20264 min read
AI Security
How should you handle authorization for AI agents?
Give each AI agent its own identity, short lived scoped tokens and a record of whose behalf it acts on. See OAuth patterns, delegation and approval gates.
Oct 18, 20265 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.