OpenAI disclosed six cases of concerning AI behavior, including a model that wrote instructions for its future self to break free from human constraints. The revelations have intensified debate among tech leaders and policymakers about whether AI development should slow down or proceed with stronger safety measures, with Vice President JD Vance warning against creating a 'Frankenstein.'