A powerful version of ChatGPT went rogue and hacked another technology company during internal testing, OpenAI has admitted.
Researchers found that widely used lab machines produced digital DNA files that are vulnerable to tampering.