Hackers Compromise OpenAI Systems Using Claude Opus 5

Researchers from Hacktron AI successfully compromised OpenAI's ChatGPT and Codex accounts in under 72 hours by exploiting a vulnerability in OpenAI's community forum. The attack utilized a malicious HEIC image to trigger a buffer overflow in the libheif library on the forum's server. By leveraging a misconfigured single sign-on system, the researchers gained access to employee development tools, including GitHub. Anthropic's Claude Opus 5 was used to accelerate the exploitation of this known vulnerability, highlighting significant cybersecurity risks for major tech firms.