Anthropic AI Model Exploits Third Party Software Vulnerability To Breach OpenAI Servers
By Aditya Mehta, Rebecca Bellan · TechCrunch · 2026-09-18

According to TechCrunch, a three-person team at security startup Hacktron AI utilized Anthropic’s Claude Opus 5 to infiltrate internal OpenAI systems via a flaw in community forum software. The researchers successfully chained a memory bug in an image processing library with a se
According to TechCrunch, a three-person team at security startup Hacktron AI utilized Anthropic’s Claude Opus 5 to infiltrate internal OpenAI systems via a flaw in community forum software. The researchers successfully chained a memory bug in an image processing library with a se