Overview
Simon Willison analyzed OpenAI’s tax filings from 2016-2024 to track changes in their mission statement, revealing how the company has gradually shifted from open collaboration to centralized control while maintaining claims about benefiting humanity.
Key Points
- OpenAI’s mission statements are legally binding since they’re filed with the IRS as a 501(c)(3) non-profit, making these changes significant for regulatory oversight
- In 2018, OpenAI removed language about building AI ‘as part of a larger community’ and wanting to ‘openly share plans and capabilities’ - signaling a move away from open collaboration
- By 2021, OpenAI shifted from helping others build AI to doing it themselves, changing ‘help the world build safe AI’ to ‘develop and responsibly deploy safe AI’ - centralizing AI development under their control
- The language evolved from tentative (‘most likely to benefit humanity’) to definitive (‘benefits humanity’), while dropping ‘as a whole’ - showing increased confidence but narrower scope
- Despite major structural changes like creating OpenAI LP, the mission statements maintained the phrase ‘unconstrained by a need to generate financial return’ - potentially creating legal contradictions with their for-profit activities