<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0"><channel><title>kosa8 blog</title><link>https://kosa8.com/blog/</link><description>Daily field notes on AI agent security.</description><language>en</language><lastBuildDate>Wed, 14 Oct 2026 12:24:36 +0000</lastBuildDate><item><title>AI agents in banks &amp; fintech: securing fraud, code and data</title><link>https://kosa8.com/blog/ai-agents-in-banks-fintech-securing-fraud-code-and-data/</link><guid>https://kosa8.com/blog/ai-agents-in-banks-fintech-securing-fraud-code-and-data/</guid><pubDate>Wed, 14 Oct 2026 07:00:00 +0000</pubDate><description>A CISO guide to the real risks of AI agents in banking, practical controls, and how kosa8’s sealed microVMs help meet PCI, SOX, DORA and FFIEC expectations.</description></item><item><title>A public LLM wrote a Chrome exploit in six days – Your agents can too</title><link>https://kosa8.com/blog/a-public-llm-wrote-a-chrome-exploit-in-six-days-your-agents-can/</link><guid>https://kosa8.com/blog/a-public-llm-wrote-a-chrome-exploit-in-six-days-your-agents-can/</guid><pubDate>Wed, 14 Oct 2026 07:00:00 +0000</pubDate><description>LLM built a Chrome exploit in six days, an AI‑driven OTP bypass appeared, and tiny models could self‑replicate. Get practical steps to secure agents.</description></item></channel></rss>
