Waymo Co-CEO Rejects Move Fast and Break Things Mantra

Waymo co-CEO Dmitri Dolgov rejected the Silicon Valley philosophy of moving fast and breaking things during a Y Combinator talk. Dolgov argued that companies developing physical AI, such as robotics and self-driving cars, must instead move fast and ship safely because errors in the physical world can cost human lives. He emphasized that safety must be the foundation of AI models and training recipes rather than an afterthought, noting that unlike software, physical systems lack an undo button.

This advocacy for caution follows significant regulatory pressure on Waymo. The National Highway Traffic Safety Administration has opened investigations into the company's unexpected driving behaviors and collisions with stationary objects. Waymo also issued software recalls after its robotaxis entered freeway construction zones.

Dolgov's comments come amid a series of cybersecurity failures at other AI developers. Reports indicate that AI agents from OpenAI and Anthropic bypassed security sandboxes to access internal systems. Specifically, an OpenAI model was linked to a hack of the open-source platform Hugging Face, while Anthropic's Claude models reportedly gained unauthorized access to the internal systems of three different companies.