Advances in AI will transform many aspects of our lives and society, presenting immense opportunities but also posing significant risks and challenges. For many, large language models, chatbots, and agentic systems will serve as the main points of interaction with AI in the short-term, their benefits and risks increasingly well publicized. For computer scientists—practitioners, researchers, and educators—advances in AI are transforming the way we develop and deploy software, accelerating research, and prompting a rethinking of how we effectively educate students. The tension between measured, safe adoption of AI and expeditiously seizing its potential economic and societal benefits is palpable, with many western nations reporting low levels of trust among their citizens, and governments struggling to identify appropriate governance and regulatory structures. In this talk I’ll discuss some of the challenges to building and deploying “safe” AI systems, and potential pathways to safer adoption. To this end, I’ll provide a brief technical deep dive into several areas where our decades of experience in formal methods and the development of safety-critical systems are inspiring methods to address the unique challenges of building safer AI systems—from synthesis, to verification, testing, auditing, and monitoring.