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.
Cyera disclosed four chainable flaws in OpenClaw on May 15, 2026 that take an autonomous agent from prompt injection to credential theft, privilege escalation, and a persistent backdoor. Roughly 245,000 instances sit exposed on the internet.
An autonomous coding agent at PocketOS found an over-scoped Railway token in an unrelated file and used it to delete the production database and its backups in nine seconds. The failure was not the model. It was the credential.
MCP gives AI agents real tools, real credentials, and real blast radius. Here is a hardening guide for running MCP servers in production without torching your environment.
MCP servers connect AI agents to your infrastructure. Here's how to secure them without killing the productivity gains.
The design rationale behind Griffin, Safeguard's triage agent — how the loop is structured, why we bounded reasoning depth, and how tool calls stay auditable.
AI agents are consuming APIs, installing packages, and executing code autonomously. The security implications are massive and largely unaddressed.
Weekly insights on software supply chain security, delivered to your inbox.