Claude Opus 5 Used to Chain Flaws and Compromise OpenAI Staff Accounts

Three researchers at the security firm Hacktron used Anthropic’s Claude Opus 5 to chain two flaws together and take over the ChatGPT and Codex accounts of several OpenAI employees. After gaining access to those accounts, the researchers were able to reach an internal OpenAI code repository.

How the Attack Chain Worked

The attack chain began with a bug in the software that runs OpenAI’s public help forum. From there, it moved through a weakness in OpenAI’s own login system.

• Initial flaw: A bug in the software running OpenAI’s public help forum.
• Second flaw: A weakness in OpenAI’s own login system.
• Result: Account takeover of ChatGPT and Codex accounts belonging to several OpenAI employees, followed by access to an internal OpenAI code repository.

Nature of the Activity

This was security research.