Every few generations, a new technology comes along that changes everything.
Imagine a remote American town in the 1920s as electricity arrived.
Before the transmission lines came, daily life was constrained by physical limits: hauling water, washing clothes by hand, preserving food with ice, with most activities ceasing after sunset. Electricity didn't transform every home overnight, and its benefits were not equally shared by everyone at first.
But as access became widespread, ordinary life quietly transformed. Electric lights extended the usable day; electric pumps, household appliances, and refrigerators lifted much of the burden of daily chores; radios brought news, music, and emotional connections from hundreds of miles away into homes and public spaces.
Electricity's initial purpose was to make life more convenient, but its deeper impact came from the new possibilities it created as more people could use it. Over time, new possibilities kept emerging—machines and computers dramatically accelerated progress in medicine, engineering, and countless other fields.
By the end of the 20th century, average human life expectancy had increased by over 20 years, and median income, adjusted for inflation, had roughly tripled.
These achievements were largely driven by improvements in healthcare, sanitation, and standards of living, many of which were enabled or accelerated by the widespread availability of electricity and related technological progress.
AI is poised to replay this pattern. AI is reaching remarkable capabilities. But the focus isn't on the technology itself, it's on what people can do with it.
It can help people understand a medical bill, learn a new skill, start a small business, care for an aging parent, make sense of a legal or financial decision, turn an idea into reality, or even advance a scientific discovery.
The novelty of electric light at night might fade quickly, but what people decide to do with that light does not. And because technology has been a reliable pathway to prosperity, we believe AI should be accessible to everyone, allowing each person to use it as they need, wherever they are, and in whatever way they choose.
But such a future will not happen on its own.
Transformative technology can concentrate power or disperse it; it can make life more convenient for a few or expand opportunity for the many. Our philosophy is based on a conviction: AI should serve people—helping them pursue their goals, augmenting their capabilities, and spreading the benefits of this technology as widely as possible.
Our first commitment is to build AI that serves humanity. This means we aim to empower broadly, not concentrate power in the hands of a few companies, governments, or individuals.
We believe a safer future is one where power is widely distributed, enabling more people in the world to participate in building a resilient ecosystem.
We are optimistic about AI because we believe it can expand human capability and well-being.
But we are also sober about the risks. Powerful systems must be safe, must be aligned with human intent, and must remain under human oversight and control. OpenAI's mission is to ensure that AGI benefits all of humanity. This means building systems that help people do more of what they choose to do themselves, not replace them.
Full automation of everything is not the future we want.
That would be bleak and dangerous. AI should help people be more, not less. As AI systems become more capable, the human role becomes more important: setting direction, making trade-offs, exercising judgment, and bringing values, taste, care, and accountability into the work.
In the future, a core human mission will be deciding what is worth doing.
We believe that, in the coming years, AI conducting AI research will be a key determinant of the pace of progress.
This is crucial because 'alignment' itself is a deep research problem. To make rapid and substantive progress, our researchers will need AI systems to test ideas, find mistakes, explore alternatives, and iterate alongside us.
However, faster technological progress makes human judgment and public collaboration more important, not less. The future should be shaped by people, institutions, and societies collectively, not only by the companies building the most powerful systems.
As frontier AI development continues, we expect coordination at national and global levels to become increasingly important.
We have long believed that ultimately, an international organization should be established to coordinate major AI efforts and collectively mitigate catastrophic risks.
Establishing cooperation and common safety as a norm is also a critical part of the path forward—especially because the problems posed by commercial and national competition are hard to avoid.
One goal of such an organization should be to enable the world to act in a coordinated manner, including, when necessary, slowing frontier development to allow societal resilience, safety, and alignment capabilities to keep pace with technical progress.
Currently, OpenAI has three core goals:
Build automated AI researchers—an AI system capable of accelerating and increasingly automating the research process itself, while remaining steerable, accountable, and deeply connected to humans.
Our internal estimation is that by March 2028, a significant portion of our research may be conducted by AI systems working in collaboration with our own researchers.
To make sufficient progress on alignment research, we believe we need AI to iterate with us. This will help us collectively navigate the transition to a 'post-AGI world,' allowing us to collectively decide the path forward.
Accelerate economic progress—by accelerating scientific advancement, raising productivity, and driving economic growth, while committing to ensuring the benefits are shared widely.
Everyone should have the opportunity to share meaningfully in the prosperity AI creates.
Provide a personal AGI for every person on the planet—empowering them to benefit from one of humanity's most transformative technologies in ways they choose.
To achieve all this, we are entering the third phase of OpenAI.
OpenAI's first phase was research-focused around AGI.
The second phase began as our research started connecting with the real world, and we became a product company: deploying our systems, learning from how people use them, and continuing to advance toward AGI safely and in alignment with our mission.
Now, we are entering the third phase.
The economy is being reshaped around AI. The central questions have become: How do we make advanced AI abundant, affordable, safe, useful, and usable enough that every person and organization can benefit?
Frontier capabilities are only part of the work; the larger task is turning that capability into tools people can actually use to create and build.
Most importantly, we believe widespread distribution of power will lead to a better future. Human history shows that concentrated power makes societies brittle, while broadly shared power makes them more resilient, adaptive, and free.
This is why ubiquity matters, and why safety, privacy, affordability, an open ecosystem, and public oversight matter.
A good AI future is not one where a few institutions hold most of the capability and benefit.
It is one where many people, numerous businesses, communities everywhere, and countries around the world can build, benefit, and hold power. We believe this transformation should belong to everyone.
If we get this right, AI can become a cornerstone of increased productivity, creativity, scientific progress, and economic opportunity for the many, and we will achieve our mission: ensuring that AGI benefits all of humanity.
