The Coding-Agent Trap: When a "Free" LLM Endpoint Is the Adversary, (Mon, Aug 31

I identified an attacker using a semi-autonomous coding agent to run an offensive operation: finding poorly secured LLM resale gateways, acquiring API access through ordinary web

<h2> No jailbreak. No exploit. No alert fired. Just a conversation. </h2> <p>In September 2025, a Chinese state-sponsored threat group ran a cyberattack against 30 organizations

<!-- SC_OFF --><div class="md"><p>One recurring issue with domain-specific agents is that overly defensive refusal behavior can make them much less useful once the workflow gets

<h2> The Problem </h2> <p>You know what's funny about "free" AI coding tools?</p> <p><strong>They're not free.</strong></p> <p>Codex CLI says it uses your OpenAI credits — and

<p>If you're building AI agents in production, you've probably wondered: <em>what's actually happening at runtime?</em> We spent six months finding out — and what we found changed

<h2> The Setup </h2> <p>MCP (Model Context Protocol) lets AI agents call external tools. Instead of just reading files and running bash, the agent gets structured access to APIs,

<!-- SC_OFF --><div class="md"><p>Sentinel Gateway is an agent-agnostic platform with its own native, Claude-based agent, designed to combine control, flexibility, and security in

<!-- SC_OFF --><div class="md"><p>Been working with agents that can execute tools (shell, APIs, etc.) and noticed a gap:</p> <p>Unsafe behavior doesn’t show up in the prompt or