(AI) Security is about Helping Everyone Else
The actual problem: enabling teams to use AI securely at the speed they operate.
Read more
Insights on
LLM Security
Research, tutorials, and updates from the Lab Rat security team.
The AI Security Management Gap: From Fog to Secure-by-Default Deployment
Most CISOs cannot clearly define 'AI security' or articulate how to manage it. The path from this fog to secure-by-default AI deployment starts with understanding where security management breaks down.
Read moreDefense-in-Depth for AI Agents: Why Input Security Isn't Enough
Defense-in-depth for AI agents requires securing both input and output layers.
Read moreThe Lopsided Risk Dynamic of AI Agent Deployments
AI agent deployments are creating the most lopsided risk dynamic in modern security.
Read more86% of Organizations Delayed AI Deployments Due to Security
But the ones who ship anyway are learning painful lessons. Here's what's actually breaking.
Read moreThe SIEM Blind Spot: Why Traditional Security Tools Miss LLM Attacks
When Salesforce Agentforce was compromised, traditional security tools saw valid HTTPS requests. The prompt injection exfiltrating CRM data went undetected.
Read moreThe Security Problem Every SaaS Company Has But Won't Talk About
AI assistants shipped fast to stay competitive. Now companies are quietly backfilling the security controls that should have been there from day one.
Read moreLangChain Serialization Vulnerabilities: When LLMs Generate Exploits
Two critical vulnerabilities show how LLM-generated content can bypass traditional security controls and execute malicious code.
Read moreThe FAFO Framework: Fast Adoption, Future Accountability
Most companies adopted the same AI security strategy: ship features fast, figure out security later. Here's why that doesn't end well.
Read moreWhy Traditional Security Tools Can't Catch LLM Attacks
Traditional security looks for malformed syntax. LLM attacks exploit meaning through natural language. It's a fundamentally different attack surface.
Read more