OpenAI forms team to study 'catastrophic' AI risks, including nuclear threats | TechCrunch