Our Principles

OpenAI, wary of a future in which AGI becomes concentrated in a small number of companies, laid out a direction for returning AI to more people. The company said that while AI could greatly expand individual capability and autonomy, the choices made now could also determine the power structures of the future.

The core principles are five: Democratization, Empowerment, Universal prosperity, Resilience, and Adaptability. First, to prevent decision-making power over AI from concentrating in just a few labs, OpenAI emphasized democratic processes and principles of equality. Second, the company said it will design products so that users can accomplish more work and create greater value.

Third, OpenAI said that new economic models and large-scale AI infrastructure are needed to achieve universal prosperity that benefits everyone. To that end, it explained it is preemptively securing compute at a scale large relative to its current revenue, and pursuing vertical integration and data center expansion to lower costs.

Fourth, on resilience, OpenAI stressed the need to address risks such as models being misused to create pathogens or carry out cyberattacks. OpenAI said it will work with governments, international organizations, and other AGI efforts to build society-wide defenses, and will use its models to help protect vulnerable open-source software and critical infrastructure.

Fifth, through adaptability, OpenAI said it will continue to revise its principles to match an uncertain future. Citing its past hesitation over releasing the GPT-2 weights and its iterative deployment strategy, the company summarized that it will expand capabilities carefully going forward and learn together with society.