Blog
Security insights, use cases, and technical deep dives from the Netallion AI Assurance team.
Introducing the AI Assurance Score: Your AI Security Posture in One Number
A single score that tells you how secure your AI usage is — across secrets, prompts, agents, compliance, and remediation. Here is how we compute it and what it means for your team.
19 Detection Rules for Agentic AI: How Runtime Defense Catches What Firewalls Miss
Prompt injection, tool misuse, data exfiltration — agentic AI creates attack surfaces that traditional security tools cannot see. Runtime Defense watches every AI interaction in real time.
Why We Chose Keycloak: Enterprise SSO for AI Security Platforms
Moving from legacy JWT to Keycloak OIDC gave us RS256 tokens, centralized MFA, and marketplace-compatible auth — all without breaking the Microsoft Marketplace purchase flow.
Secrets Are Leaking Through Slack, Teams, and Jira — and Nobody Is Scanning
Collaboration tools are one of the largest unmonitored surfaces for secret exposure. Developers share credentials in Slack threads, Teams chats, and Jira tickets every day.
Prompt DLP: Stop Secrets and PII Reaching LLM Providers
Developers paste credentials into ChatGPT, Copilot, and Claude every day. Prompt DLP detects secrets and PII in outbound AI prompts before they leave your enterprise boundary.
One-Click Remediation: From Detection to Secret Rotation in Seconds
Finding secrets is only half the problem. Close the loop with one-click rotation into Azure Key Vault, GitHub token revocation, and AWS key deactivation — with full audit trail.
MCP Governance and the Rise of Agentic AI Security
AI agents are calling external tools via MCP servers with no inventory, no trust scoring, and no governance. Here is how to bring visibility and control to agentic AI environments.
The Azure Monitor Blind Spot: Why Your Logs Are Leaking Secrets
No mainstream security tool scans Azure Monitor logs for secrets. We analyzed why this gap exists and what it means for organizations running on Azure.
BPE Tokenization for Secret Detection: 98.6% Recall vs. 70.4% for Entropy
How byte-pair encoding tokenization analysis outperforms traditional entropy-based approaches for detecting generic secrets in code and logs.
The NHI Problem: 100 Non-Human Identities for Every Developer
Non-human identities outnumber developers 100:1 in the average enterprise. Most security teams have no inventory of them. Here is how to fix that.