OpenAI co-founder John Schulman exits Anthropic after six months

Top AI researchers are moving faster than the models they train.

Just six months after leaving OpenAI to join rival Anthropic, OpenAI co-founder John Schulman has already departed the company.

The exit was first reported by The Information, with Schulman later confirming the move on X. Schulman noted that leaving wasn't easy given the stimulating environment and talented team at Anthropic, but said he decided to pursue another opportunity that he found extremely compelling.

Anthropic Chief Science Officer Jared Kaplan confirmed the departure in a statement, saying the firm was sad to see him go but fully supported his decision to pursue new opportunities.

A brief pitstop in the alignment wars

Schulman was an OpenAI veteran of nine years and a key architect behind the reinforcement learning techniques that powered ChatGPT.

When he departed OpenAI in August, he explained in an internal note that his goal was to deepen his focus on AI alignment and return to hands-on technical research. At the time, Anthropic looked like the ultimate safe haven for safety-minded researchers.

His exit last year was part of a broader talent exodus from OpenAI, which saw the departures of Chief Scientist Ilya Sutskever, CTO Mira Murati, and Superalignment co-lead Jan Leike, who also joined Anthropic last May.

Anthropic itself was founded in 2020 by former OpenAI executives Dario and Daniela Amodei around the promise of safety-first AI development. According to reports, at least 23 former OpenAI staffers have left the firm to start their own companies.

Why it matters

We think Schulman's rapid exit highlights an intense new reality in the frontier AI talent race: even the most prestigious labs can barely keep top minds locked down for a single product cycle.

While Schulman hasn't revealed his next move yet, it looks increasingly likely that he will follow his former OpenAI peers down the entrepreneurial route. With founders like Sutskever and Murati commanding massive valuations for new ventures, someone with Schulman's hands-on alignment credentials would have no trouble securing massive compute and backing on day one.

For Anthropic, this departure is a subtle reputational bump. The company positioned itself as the premier sanctuary for researchers who felt OpenAI was moving too fast toward commercialization. Losing a co-founder of Schulman's stature after just six months suggests that internal dynamics, alignment philosophy, or resource competition in frontier labs are far more complex than simple corporate marketing suggests.

For developers and the broader AI ecosystem, it is a reminder that the intellectual center of gravity is fragmenting away from the original mega-labs and dispersing into a new wave of focused, highly funded ventures.