The man who co-founded OpenAI to save humanity from reckless AGI wanted to run it like a family business, with his kids as heirs.
The Summary
- Sam Altman testified he was "extremely uncomfortable" with Elon Musk demanding complete control over a proposed for-profit OpenAI subsidiary in 2017
- Altman described the conversation as "particularly hair-raising", revealing Musk floated handing OpenAI to his children
- The testimony exposes the core tension that split OpenAI: Musk wanted absolute control over the thing he co-founded to prevent absolute control by anyone
The Signal
Sam Altman's courtroom testimony pulls back the curtain on OpenAI's 2017 fracture point. Musk, who helped launch OpenAI as a nonprofit safety project, insisted on total control over a new for-profit arm. Not board oversight. Not shared governance. Complete control. Then, according to Altman's account, Musk suggested he might hand the organization to his children.
This wasn't a casual what-if. This was the ideological schism that would define AI development for the next decade. Musk argued you can't build AGI safely through a committee. Altman argued you can't build it safely if one person holds the keys.
"The man who warned loudest about AI risk wanted to concentrate that risk in his own hands."
Key contradictions:
- OpenAI was founded explicitly to decentralize AI power
- Musk's safety concerns drove the nonprofit structure
- His proposed solution was maximum centralization under his control
The irony compounds when you map it forward. Musk left OpenAI in 2018, started xAI in 2023, and now runs an AI company with zero external governance. OpenAI converted to a capped-profit structure without him. Both organizations claim the safety mantle. Both rejected the other's model as too dangerous.
Altman's "extremely uncomfortable" phrasing is load-bearing. This wasn't a polite disagreement about org structure. This was a fundamentally incompatible view of how you govern the most powerful technology humans have built. Do you distribute control and risk moving slowly, or concentrate it and risk moving recklessly?
Seven years later, we're living in the answer. OpenAI raised billions from Microsoft, shipped ChatGPT, and sparked the global AI race. Musk built xAI, launched Grok, and integrated it into X. The decentralized safety approach and the singular-vision approach both exist. Neither prevented the acceleration both founders feared.
The Implication
Watch how AI labs structure power as capabilities scale. The governance question Musk and Altman fought over in 2017 isn't resolved. It's just distributed across a dozen companies, each making the same choice. Google DeepMind, Anthropic, xAI, OpenAI are all led by individuals or small teams with extraordinary control over model development and deployment.
The real test comes when one of them reaches AGI, whatever that means. The organizational structure chosen today determines who makes the call tomorrow. Altman won the battle for OpenAI's governance. But Musk's question still stands: can you build something this powerful without someone in charge who can act fast when it matters? We're about to find out.