Quoting Sam Altman

Court filings have a way of surfacing the quiet parts companies usually keep quiet, and this one is a doozy. In an October 1, 2022 email to OpenAI's board, Sam Altman floated releasing a GPT-3-level model that could run on ordinary consumer hardware. Not because it would help researchers or hobbyists, he wrote, but because doing it first would make life harder for competitors trying to raise money for similar projects.

The email, unearthed as part of the Musk v. Altman litigation, names Stability AI directly as the company OpenAI wanted to beat to the punch. The logic Altman laid out is blunt: ship something roughly as capable before someone else does, and you starve the next wave of open-model startups of the funding and attention they need to get off the ground. It reads less like an open-source mission statement and more like a chess move against the field.

This matters because OpenAI has spent years positioning itself as the responsible, safety-first lab while smaller labs get accused of moving recklessly. An email like this complicates that story. It suggests that at least some internal thinking around 'openness' in late 2022, right before ChatGPT's launch reshaped the entire industry, was tactical rather than philanthropic. Releasing a capable-but-not-frontier model wasn't about democratizing AI. It was about controlling who else got to play.

There's also a timing wrinkle worth sitting with. This was written weeks before ChatGPT went live and turned OpenAI into a household name almost overnight. The board was apparently weighing open-source strategy as a competitive lever at the exact moment the company was about to stop needing that lever altogether. Whether the GPT-3-sized local model ever shipped in the form Altman described is a separate question; what the email reveals is the mindset behind decisions that later got dressed up in safety language.

Musk's lawsuit has already forced a lot of internal OpenAI correspondence into daylight, and each new batch tends to reframe some part of the company's public narrative. This one is a small document with a big implication: statements about openness and safety from major AI labs are worth checking against the emails nobody expected to become public.