Notes from our work in AI security: what breaks, why it breaks, and what we built to fix it.
Every SOC analyst has opened VirusTotal, GreyNoise, and AbuseIPDB in three tabs for the same IP. Here is the open-source MCP server we built to collapse those three tabs into one tool call, one normalized verdict, and one answer.
READ ARTICLE
CMMC readiness efforts rarely fail at the control mapping stage. They fail at evidence: scattered artifacts, stale policies, and gaps nobody catches until week six. Here is what we built to fix the operational layer underneath the checklist.
The OWASP Top 10 for LLM Applications 2026 was published in August 2026. Here is the list, what changed, and how each risk lines up with the NIST AI RMF, based on the mapping published with the release.
We run multiple AI agents in production every day. Here are the general lessons that changed how we secure agentic systems: scoped identities, untrusted inputs, human approval for irreversible actions, and logs you can act on.
