<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
	<channel>
		<title>Sandboxing on Security Architects</title>
		<link>https://securityarchitects.com/tags/sandboxing/</link>
		<description>Recent content in Sandboxing on Security Architects</description>
		<generator>Hugo</generator>
		<language>en-us</language>
		
		
		
		
			<lastBuildDate>Sat, 18 Jul 2026 10:00:00 -0700</lastBuildDate>
		
			<atom:link href="https://securityarchitects.com/tags/sandboxing/index.xml" rel="self" type="application/rss+xml" />
			<item>
				<title>Tens of thousands of exposed AI agents — sandboxing must be the default</title>
				<link>https://securityarchitects.com/post/sandbox-by-default/</link>
				<pubDate>Sat, 18 Jul 2026 10:00:00 -0700</pubDate>
				<guid>https://securityarchitects.com/post/sandbox-by-default/</guid>
				<description>&lt;p&gt;The always-on personal AI agent is the most exciting thing happening in&#xA;computing right now — and the most exposed. Between late January and early&#xA;February of this year, &lt;a href=&#34;https://www.bitsight.com/blog/openclaw-ai-security-risks-exposed-instances&#34;&gt;Bitsight&amp;rsquo;s internet-scale scanning&lt;/a&gt;&#xA;observed &lt;strong&gt;more than 31,000 distinct exposed instances&lt;/strong&gt; of the most popular&#xA;personal-agent stack, listening on its default port across technology,&#xA;healthcare, finance, government, and insurance networks. Independent&#xA;scanning teams reported even larger numbers, including thousands of&#xA;instances vulnerable to remote code execution.&lt;/p&gt;</description>
			</item>
	</channel>
</rss>
