xAI co-founder Igor Babuschkin departs to launch safety fund

Igor Babuschkin, the technical heavyweight who helped Elon Musk build xAI from scratch, has officially called it quits.

On Wednesday, Babuschkin posted on X that it was his last day at the company he co-founded back in 2023. As head of engineering, he helped scale xAI into an elite frontier lab capable of trading blows with OpenAI, Anthropic, and Google DeepMind in just a couple of years.

Now, he is launching his own firm, Babuschkin Ventures, focused on backing AI safety research and supporting startups trying to advance humanity and unlock mysteries of the universe. He credited a dinner conversation with Max Tegmark, founder of the Future of Life Institute, for inspiring the pivot.

From StarCraft to the Memphis sprint

Babuschkin has quietly been at the center of the modern AI wave.

Before launching xAI, he was at Google DeepMind working on AlphaStar, the AI that famously dismantled world champion StarCraft II players in 2019. He then spent years at OpenAI as a core researcher before ChatGPT ever hit the public.

When he teamed up with Musk in 2023, the goal was to build a different kind of AI lab. Babuschkin noted that Musk taught him two massive lessons: dive headfirst into messy technical weeds yourself, and operate with a maniacal sense of urgency.

That urgency showed up most clearly in Memphis, Tennessee, where xAI assembled its flagship supercomputer in just three months. Industry veterans had called the timeline impossible. The project got done, though it also drew sharp criticism from environmentalists over temporary gas turbines pumping emissions into nearby neighborhoods.

State-of-the-art models, sloppy guardrails

Babuschkin's departure lands right in the middle of a messy operational stretch for xAI.

While the company's frontier models consistently post top-tier benchmark scores, Grok has suffered a relentless string of self-inflicted PR disasters over the past few months.

The chatbot was caught surfacing Musk's personal X posts as definitive answers to controversial topics. In another incident, it went on an antisemitic rant and labeled itself Mechahitler. Most recently, xAI faced backlash over Grok features enabling users to churn out non-consensual deepfake nudes of celebrities like Taylor Swift.

Despite the chaos, Babuschkin struck a warm tone on his exit, comparing the feeling to a proud parent driving away after dropping their kid off at college.

Why it matters

Losing an engineering pillar is a major hit for xAI at a time when frontier talent is the scarcest currency in Silicon Valley.

Babuschkin was the bridge between DeepMind-grade research and Musk's hyper-aggressive infrastructure buildout. Keeping that momentum without him will test whether xAI's engineering culture can endure outside Musk's personal gravity.

We also think the contrast in Babuschkin's departure is telling. He is trading high-stakes model shipping for venture capital centered explicitly on AI safety. It highlights a widening rift across the ecosystem: while labs race to deploy raw intelligence at breakneck speed, the guardrails and societal impact are falling dangerously behind.

Grok's repeated misfires show what happens when shipping velocity completely outpaces alignment work. If one of the chief architects behind the machine feels the next logical move is funding safety from the outside, the rest of the industry ought to be taking notes.