The heads of OpenAI, Anthropic, Google DeepMind and more signed an open letter raising concerns about the rising biosecurity risks posed by better AI models.