What Datadog Learned About AI Agent Security by Monitoring Its Own Agents
Datadog's internal agent monitoring revealed that 69% of input tokens live in system prompts teams can't inspect, coding agents execute repo-controlled code before the first user message, and prompt injection bypasses model evaluation entirely.
How Datadog Saves $1 Million a Month on AI: The Optimization Playbook
Datadog cut over $1M/month in AI costs by migrating models, adjusting effort levels, and deploying context optimization. The specific numbers and methods behind each lever.
JFrog's Agent Guard Targets the MCP Servers Nobody in Security Approved
36% of MCP servers and agent skills tested carry a security flaw. JFrog's Agent Guard adds an approved-only proxy so developers can't connect coding agents to ungoverned tools.
AI Coding Speed Is Outrunning Secrets Security
Nearly 30 million secrets leaked on GitHub in 2025 (up 34% YoY). AI-assisted code leaks at nearly 2x the rate. What DevOps security teams should do now.
JFrog's AI Asset Detection: Why Shadow AI Is the New Shadow IT
Coding agents are pulling MCP servers, models, and AI-generated code into your pipelines without approval. JFrog's detection framework classifies four categories of AI assets and explains how to build a continuous inventory before they become supply chain risks.
JFrog's Package Traffic Controller Closes the Last Open Door in Supply Chain Security
AI agents bypass artifact repos entirely. JFrog's new network-layer tool intercepts every outbound package request — developers, agents, and everyone else — and routes it through curation policy.
LiteLLM Compromised: How the AI Gateway Became the Attack Surface
On March 24, attackers compromised LiteLLM via a single unpinned CI dependency, deploying credential harvesters and Kubernetes backdoors across affected environments.