An experiment by the InsideAI has raised fresh concerns about artificial intelligence safety after showing that a humanoid robot powered by ChatGPT could be persuaded to bypass its safeguards and fire ...