Found a bug in https://beta.openai.com/codex-javascript-sandbox where a simple request get's the model to keep spitting out random thing. I added a white cover cause the data contained auth tokens and other potentially sensitive data.
I believe it's a bug that allows the bypassing of the token limits since it keeps going till 20K tokens are reached, then if I refresh and send it again it goes on another 20k.
Is there any bug bounty program? I emailed a week ago asking support about it and about a consent for testing for such bugs and trying cybersecurity related things with the various models but got nothing.