An independent security research team gained access to the internal software repository of OpenAI by exploiting a vulnerability with an employee's ChatGPT account. Hacktron researchers used Anthropic's Claude software to execute the breach, which allowed them to read and propose changes to the company's private cache of code.

OpenAI provided a $6,500 bounty to the bug-hunting team after they reported the flaw. The incident exposed the company's algorithmic secrets and the private repository used to store its internal software.