In a blog post disclosing the incidents, OpenAI said that, in the first case, an agent powered by GPT Sol 6.1 disabled automatic alerts after a team...