- Why “change architecture”?
- Why does PBC mode matter?
- Key elements of structural adjustment
- # Sam Altman full text of open letter to staff and stakeholders (Chinese translation)
- Now, we think it’s time to restructure. We have three goals:
OpenAI announced that its organizational structure would change significantly: OpenAI for-profit LLC, controlled by a non-profit organization, would be transformed into Public Benefit Corporation, PBC. This structure is intended to ensure that AGI (general artificial intelligence) benefits all of humanity while preserving mission orientation and introducing more flexible capital structures and resource support to better advance its core mission . ** The new structure highlights the following:
-
** Non-profit organizations continue to have absolute control** and to be the main shareholders in PBC;
-
The need for the PBC to balance public interest and corporate mission with profitability;
-
To abandon the old “capture profit model” (apped-profit) and replace it with a regular equity structure;
-
Access to greater financing capacity to support global computing resources and modelling training;
-
Continuous focus on the development of AI security, democratization applications and open-source models.
Sam Altman (OpenAI CEO) in his letter to the staff stressed: OpenAI has never been a “normal company” whose mission is not to maximize profits, but to maximize the positive impact on human beings. AGI should not serve only a few, but should empower everyone. AGI’s vision has shifted from a “mysterious tool” to a “manufacturing tool for all”:
-
Early fears that AGI was too powerful to be controlled by a small number of “trustworthys”;
-
Now, OpenAI’s vision is to turn AI into a “human tool” for real scenarios such as scientific research, medical advice, productivity enhancement, etc.;
-
The concept of “AGI** for democratization” is the direction that OpenAI continues to follow.
ChhatGPT has shown direct influence:
-
Extensive use in learning, science, writing, program development, etc.;
-
User demand is far greater than the current supply and AI calculates resource costs to be extremely high;
-
OpenAI wants to expand its service capacity to meet “globalization needs”.
The restructuring was driven by three core objectives: ** Sustainability and financing: access to hundreds of billions of dollars in resources for training and deployment of stronger AI (which may reach trillion levels in the future); ** Mission expansion: making non-profit organizations the most influential public benefit agency in history; ** Promotion of security, alignment, democratic AI development**: adherence to values, opposition to autocratic AI, continuous promotion of mechanisms such as testing of Red Brigades, transparency of models, etc. Sam Altman also promised, “We want to open up some powerful models.”
Why “change architecture”?
OpenAI was originally initiated by a non-profit organization with the aim of** ensuring that universal artificial intelligence (AGI) benefits all of humanity, rather than being monopolized by a few or businesses. In 2019, OpenAI set up a “LLC” structure to finance the enormous numeracy and financial pressure required for model training. ** Today, in the face of greater demands for arithmetic, pressure for commercialization and the need to advance the AI democratization mission, OpenAI decided to upgrade the structure again: The transformation of profit-making entities under the flag from limited liability companies (LLCs) to public interest companies (Public Benefit Corporation (PBC)) is a new business structure that requires consideration of both public and shareholder interests. This model is also used in AGI laboratories such as Anthropic, xAI and others.
Why does PBC mode matter?
**PBC = Profit + Mission ** This is a business model that balances capital efficiency with ethics and has been adopted by several leading AI companies (e.g. Anthropic, xAI). Its benefits include:
-
Financing as traditional companies;
-
The long-term impact on society and the public must also be continuously considered in operations;
-
Non-profit organizations can act as “gatekeepers” to prevent AI from departing from its original mission.
Key elements of structural adjustment
• Non-profit organizations remain “controlled”
-
OpenAI ‘ s non-profit organization will remain** under full control** of the new PBC;
-
and will be its ** largest shareholder**, ensuring that mission orientation is not influenced by market interests;
-
The decision was the result of consultations with the Attorney General ‘ s Offices of California and Delaware.
Structure changed from a “profit cap” to a “standard equity model”
-
Past LLC use capped-profit to avoid excessive returns to investors;
-
Standard equity structures will now be used to increase ownership in order to introduce funds more effectively;
-
Sam Altman stressed: “This is not a sale of companies, but a simplified structure.”
• The PBC will act as a profit-making entity, contributing to the realization of the mission
-
More sustainable and scaled-up financing capacity (for training models, deployment of products);
-
The profits will be returned to non-profit organizations in support of social missions such as global public services, education, scientific research, etc.
# Sam Altman full text of open letter to staff and stakeholders (Chinese translation)
** OpenAI is not an ordinary company, nor will it ever be.** Our mission is:** to ensure universal artificial intelligence (AGI) for the benefit of all humanity. When we started OpenAI, there was no detailed blueprint to achieve this. At first, we sat around the kitchen table and discussed what kind of research to do. At that time, we never imagined products, business models, much less that AI was being applied to medical counselling, learning, productivity enhancement, and so on. And we didn’t realize that training and deploying these models would cost billions, if not hundreds of billions of dollars, of numeracy and money. We don’t really know how AGI will be constructed, how it will be used. Many at the time imagined that AGI was a “better tool” that could provide advice to very few scientists, presidents, and so on. In the early days of OpenAI, there were really a lot of people who thought that AI should be in the hands of a small group of people who “manage it”. But now we see another possibility: **AGI will become the most powerful tool in human history, directly empowering everyone. If we can do this, we believe that people will use these tools to empower one another, to create amazing things, and to promote social progress and a better quality of life. Of course, AI cannot be used only for good intentions, but we believe that:** the power of good will goes far beyond evil. We are firmly on the path of “Democratization AI.” We want to put these extraordinary tools in everyone’s hands. We are surprised and proud that users are using the results they have created. We want to open up some very powerful models; we want to give users more freedom to decide how to use ChatGPT, even if their moral framework does not necessarily coincide with us, within legitimate and reasonable parameters. We believe that this is the best way forward: **AGI should empower everyone to benefit each other and, of course, we know that not all share that view. ** We want to build a “brain of the world” so that everyone can use it in the simplest way to do anything (as long as it does not violate the freedom of others).** People are using ChatGPT to upgrade research, write codes, learn knowledge, and cope with life’s problems. It’s the most direct expression of our mission to see that our services have brought positive value to so many people. But the user needs go far beyond our current supply. We have to set access restrictions and the system is running slowly. As the system becomes more capable, people will do more and better, and we need to make structural changes to support this trend.
Now, we think it’s time to restructure. We have three goals:
** Access to adequate resources and operational capacity, allowing our services to be truly open “to all of humanity” — which now requires hundreds of billions of dollars, possibly even trillions in the future. We believe that this is the path to the most powerful mission. ** Building the most effective and influential AI non-profit institution in history — focusing on the positive impact of highly leveraged societies: education, health, public services, etc. ** Safe and useful AGI** — We will continue to work in the direction of security and alignment: red team testing, model regulation files, transparency of the system, etc. ** We firmly believe that the democratization of AI must prevail over the centralized AI. The reason we insist on being controlled by non-profit organizations is the decision that was made after an in-depth dialogue with civil leaders, the Attorney-General’s Offices of California and Delaware. In the future, we will continue to move forward with them, with Microsoft, with our newly appointed non-profit committee. OpenAI was, is and will continue to be an organization controlled by non-profit organizations. The LLC under our flag will be transformed into a **public interest company (PBC), with the same mission. Many purpose-oriented organizations like Anthropic, xAI, Patagonia, etc., have this structure, which we believe is also applicable to us. The complex “limited profit” structure of the past is no longer applicable in the current AGI landscape. We will use simpler equity structures to give everyone a share. This is not a sale, but an optimisation of the structure. Non-profit organizations will continue to control and become important shareholders in the PBC, the proportion to be determined by an independent financial consultant. This will give non-profit institutions a stronger financial capacity to support different communities around the world to benefit from AI. We look forward to the recommendations of the non-profit committee to help us ensure that AI truly benefits everyone and promotes health, education, public services and scientific discovery. We believe that this will enable us to move forward faster and safer with AGI and truly deliver powerful AI tools to everyone. The building of AGI is a brick that we’ve laid for human progress, and we can’t wait to see the next piece that you’ll be laying. ** – Sam Altman** May 2025 Original language: https://openai.com/index/volving-our-structure/