In today’s AI Deep Dive, we explore OpenAI’s latest safety innovation—deliberative alignment—designed to make AI responses more policy-aware and secure. We also discuss xAI’s efforts to broaden access to its Grok chatbot with a new iOS app and web version. Meanwhile, Tetsuwan Scientific is revolutionizing research by combining AI and robotics to automate scientific experiments. Lastly, we unpack Palantir and Anduril’s consortium for Pentagon AI contracts and its implications for defense innovation.