Waymo's CEO say this is why he ignores Silicon Valley's most famous mantra

AI's early needs safety, not a catchy 2000s tech slogan, says the co-CEO of Waymo.

"In Silicon Valley, there's a communal mantra to 'move accelerated and break things,'" Dmitri Dolgov said astatine a Y Combinator talk released connected Monday. "However, erstwhile you're dealing pinch atoms alternatively of bits, breaking things is not really okay."

Waymo was spun retired of Google's self-driving car task successful 2016 but remains owned by Alphabet. Dolgov, who has worked connected the task since 2009, shared his type of Mark Zuckerberg's original mantra, which guided Facebook during its hyper-growth years.

Dolgov's type applies to beingness AI, wherever the stakes are high, he said.

"So the point you person to do is to move accelerated and vessel safely," he said.

He said that self-driving cars and robotics require "robust" systems to beryllium built from the get-go.

"You person to build AI models, and you person to build training recipes wherever safety is the foundation and not an afterthought, not an add-on," he said.

The Waymo co-CEO besides explained really different the costs of errors is successful integer AI and beingness AI.

"You person a connection exemplary aliases a chatbot aliases a co-pilot, and it makes a mistake, usually it costs you a retry. In the beingness world, the costs of a correction tin beryllium measured successful quality lives, not tokens," he said. "There's simply not an undo and a retry button."

Waymo has mostly avoided awesome injury-causing crashes compared to its mileage, but has faced regulatory scrutiny complete package glitches. In 2024, the National Highway Traffic Safety Administration opened an investigation into unexpected driving behaviour aft reports of collisions pinch stationary objects and postulation rule violations, prompting package recalls. In 2025 and earlier this year, Waymo issued wide recalls aft several robotaxis entered freeway building zones.

Dolgov's comments travel aft a period of awesome cybersecurity mishaps successful the ample connection exemplary world that prompted calls for much AI information regulation.

Late past month, Hugging Face, an open-source AI platform, announced that it had knowledgeable a information breach successful which an AI supplier had accessed immoderate of its soul datasets. OpenAI said its models escaped a sandbox environment, accessed the internet, and were responsible for the rogue hack.

Last week, OpenAI rival Anthropic disclosed a akin incident, saying that it recovered cases of Claude models gaining unauthorized entree to soul systems of 3 different companies.

In consequence to the OpenAI incident, Aaron Levie, the CEO of Box, said the incident was an illustration of the "wild times" AI tin create.

"If you were wondering really powerful AI is getting, Agents are now tin of escaping retired of systems, uncovering their measurement to the internet, discovering zero time information vulnerabilities on the way, and past breaking into outer systems - each successful an effort to complete their goal," he wrote connected X past month.

Follow Following

Every clip publishes a story, you’ll get an alert consecutive to your inbox!

Look retired for an alert successful your inbox the adjacent clip publishes a story!

Every clip a caller communicative is published, you’ll get an alert consecutive to your inbox!

Look retired for an alert successful your inbox the adjacent clip a caller communicative is published!

By clicking “Sign up”, you work together to person emails from Business Insider. In addition, you judge Insider’s Terms of Service and Privacy Policy.

More stories by More stories from