Thinking Machines terminates CTO Barret Zoph for 'unethical conduct,' appoints Soumith Chintala as successor

AI startup Thinking Machines has terminated Chief Technology Officer Barret Zoph effective immediately following an internal investigation into "unethical conduct," according to sources familiar with the matter. Soumith Chintala, creator of the PyTorch deep learning framework and former Meta AI research scientist, has been appointed as his replacement.

The termination follows Zoph's high-profile departure from OpenAI in late 2025, where he served as a research lead working on large language model scaling. His move to Thinking Machines was positioned as a major coup for the startup, which has been developing specialized AI hardware and software systems. Sources indicate the termination wasn't performance-related but stemmed from violations of company ethics policies, though specific details remain undisclosed.

Chintala's appointment signals a strategic shift toward open-source AI development. As founding maintainer of PyTorch at Meta, Chintala helped build what became the most widely used research framework in deep learning. His leadership suggests Thinking Machines may increase commitment to open models and developer ecosystems, contrasting with the closed approach of competitors like OpenAI and Anthropic.

This leadership shakeup coincides with OpenAI's January 14 announcement that Zoph would be returning to the company alongside researchers Luke Metz and Sam Schoenholz. OpenAI President Fidji Simo stated the move "has been in the works for several weeks," raising questions about coordination between the companies and whether Zoph's departure from Thinking Machines was anticipated.

The incident highlights persistent governance challenges in the rapidly scaling AI industry. While startups compete aggressively for top talent, this case demonstrates how quickly executive transitions can occur when ethical boundaries are crossed. Thinking Machines now faces the dual challenge of managing reputational damage while onboarding a new technical leader during a critical growth phase.

Chintala inherits a complex technical portfolio including Thinking Machines' neuromorphic computing hardware and proprietary LLM training infrastructure. His immediate priorities will likely include stabilizing engineering operations while evaluating strategic alignment between the company's hardware roadmap and software stack. Industry observers will monitor whether Chintala pushes for greater open-source contributions similar to his PyTorch stewardship at Meta.

Neither company has disclosed specifics regarding Zoph's alleged misconduct. Legal experts note that without transparency about the ethics violation, speculation could damage both Zoph's career and Thinking Machines' employer brand. The situation underscores how AI labs face heightened scrutiny beyond technical capabilities, with cultural alignment and ethical compliance becoming critical hiring filters.

Thinking Machines joins Anthropic, OpenAI, and other AI labs navigating executive turbulence amid breakneck industry growth. The resolution of this transition will test whether specialized AI hardware startups can maintain technical velocity while establishing mature governance structures expected of companies handling increasingly powerful AI systems.