Altman's open-source memo
A 2022 email surfaced from the Musk v. Altman record shows Sam Altman pitching an open-source model to OpenAI's board as a way to starve rivals of funding.
On October 1, 2022, Sam Altman emailed OpenAI's board about "open source strategy." The plan he described was to build a language model with roughly GPT-3's ability that runs on an ordinary consumer computer, and give it away for free. The reason he gave was not access or democratization. It was that releasing such a model first would "discourage others from releasing similarly-powerful models" and make it "harder for new efforts to get funded" — and that OpenAI should move before Stability AI did.
The best reason to build this and release it now is that Stability is going to do something similar. — Sam Altman, to OpenAI's board (Oct 1, 2022)
The document is dated primary evidence, pulled from the discovery record in Elon Musk's lawsuit against Altman and OpenAI. The long-running complaint that "the company called OpenAI stopped being open" is old and well-worn. What is new is the CEO, in his own words, naming a free release as an offensive move against competitors rather than a gift to the public.
The timing of its resurfacing sharpens it. The email circulated on July 20, 2026, two months after a jury dismissed Musk's claims on statute-of-limitations grounds. It has outlived the lawsuit that unearthed it, and now reads less as legal ammunition than as a plain statement of how one of the field's loudest advocates for and against openness actually weighs the choice: as a lever on rivals' balance sheets. OpenAI did eventually ship open-weight models in 2025, three years after this pitch. By then, free models had already become the standard weapon in the market it was entering.
The lenses
The facts
How this connects
Tap a node to open it