<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Ai-Gateway on CuraSec</title><link>https://curasec.metacog.co.kr/tags/ai-gateway/</link><description>Recent content in Ai-Gateway on CuraSec</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Thu, 10 Sep 2026 14:58:06 +0000</lastBuildDate><atom:link href="https://curasec.metacog.co.kr/tags/ai-gateway/index.xml" rel="self" type="application/rss+xml"/><item><title>10% of Internet-Exposed LiteLLM Gateways Use Default "sk-1234" Admin Key</title><link>https://curasec.metacog.co.kr/insights/2026-09-10-nearly-1-in-10-exposed-litellm-gateways-accepted-the-example/</link><pubDate>Thu, 10 Sep 2026 14:58:06 +0000</pubDate><guid>https://curasec.metacog.co.kr/insights/2026-09-10-nearly-1-in-10-exposed-litellm-gateways-accepted-the-example/</guid><description>&lt;ul>
&lt;li>&lt;strong>Engineer — Act:&lt;/strong> If you run LiteLLM, immediately check whether your admin key is still &amp;ldquo;sk-1234&amp;rdquo; and rotate it to a strong credential; a compromised gateway exposes all upstream model API keys and full prompt/completion history to anyone who finds the instance.&lt;/li>
&lt;li>&lt;strong>SOC/IR — Plan:&lt;/strong> Build a detection rule to flag any LiteLLM API requests authenticating with the literal string &amp;ldquo;sk-1234&amp;rdquo;, and sweep existing gateway/proxy logs since initial deployment for unauthorized admin activity.&lt;/li>
&lt;li>&lt;strong>Leader — Plan:&lt;/strong> Direct engineering to inventory all internal and vendor-managed LiteLLM deployments and confirm no default admin credentials are in use; a misconfigured AI gateway exposes both uncapped API spend and the full record of what your applications send to model providers.&lt;/li>
&lt;/ul></description></item></channel></rss>