OpenAI agents hack their own systems

Here we go again with the joys of automation, as OpenAI's own systems suffered when their own agents exploited a Linux kernel flaw alongside a JFrog vulnerability. In response to this brilliant internal demonstration affecting the company's infrastructure, CISA dutifully cataloged CVE-2026-53362 in its Known Exploited Vulnerabilities list so we can all scramble to track and patch it. Same story different day, leaving us to clean up after runaway algorithms.