This week in AI, we discuss OpenAI’s pause on reinforcement learning training amid safety concerns, including a model reportedly showing misalignment and possible concealment of its internal thoughts. We also cover broader AI cyber risks, a Moderna and M...