Sam AltmanCo-founder & CEO

It feels like both a watershed moment and the expected culmination of a decade of preparation. Inside OpenAI, we have anticipated and planned for this exponential trajectory for years, even when very few people outside believed artificial general intelligence was achievable. The difference today is that society has collectively recognized that this capability is arriving. While that realization triggers understandable fear and recalibration, our role is to help society absorb the transition constructively.

It is a surreal experience, but I try not to take it too personally. When monumental shifts occur, people naturally search for a focal point or a scapegoat to absorb their fears about the future. The creation of this technology is actually the collective achievement of brilliant scientists and engineers rather than a single individual. I recognize that anyone in this leadership seat would face identical public reactions.

Fatherhood has been the single greatest blessing of my life. It did not suddenly make me care about not destroying the world, because I was deeply committed to civilizational safety long before having children. However, it has been wonderfully clarifying in terms of time management and purpose. It sharpens your focus on what aspects of human agency, creativity, and the natural world must be preserved as this technology unfolds.

I strongly advocate for limiting technology exposure for young children across the board. Developing minds need to explore nature, interact physically, and play outside. When I see small children glued to touchscreens consuming short-form algorithmic media, my instinct is that it harms cognitive development. Children born today will undoubtedly need AI fluency as adults, but early childhood is not the appropriate stage to introduce conversational models.

Seeing an actor mirror your walk and mannerisms on the street outside your home is a bizarre, simulation-like experience. However, films prioritize dramatic storytelling over biographical accuracy, so you cannot take them too seriously. Portrayals featuring underground bunkers and firearms on walls are complete Hollywood fiction, even if they look dramatic on screen.

I maintain a practical survivalist interest and believe society systematically underinvests in resilience for low-probability, high-consequence events. Having basic food supplies, water purification capabilities, and masks on hand before crises hit is simple prudence. The global response to recent pandemics demonstrated that both institutions and individuals often fail to update their preparedness habits after experiencing a major disruption.

No, I do not have any intention of running for political office. While I once briefly considered the California governorship because of my love for the state and concerns over its administrative challenges, my experiences in technology leadership have confirmed that politics is not for me. Guiding the development of artificial general intelligence is vastly more impactful.

OpenAI will not operate like a standard public corporation. Our overarching mission will remain governed by our nonprofit board, and we will explicitly remind public shareholders that humanity's shared interests will always supersede short-term quarterly stock performance. If market expectations and safety come into conflict, we will prioritize safety and mission integrity without hesitation, even if that decision temporarily depresses our valuation.

At the time, nobody took AI as seriously as they do now, and I believed holding zero equity would clearly signal that my decisions were motivated purely by humanity rather than personal financial enrichment. In hindsight, that decision was misunderstood and created unnecessary skepticism, because people in the commercial world distrust incentives they cannot categorize. Given how the public interpreted it, taking standard equity would have avoided substantial friction.

Our track record over the past decade proves otherwise. We have voluntarily halted advanced training runs specifically to allow alignment, interpretability, and monitoring techniques to catch up with raw model capabilities. For years, we urged governments and the broader industry to prepare for these exact thresholds. Now that the broader world recognizes the gravity of the technology, we have an unprecedented window to establish meaningful shared guardrails.

The company and the broader industry are expanding at a velocity unseen in corporate history, compressing several years of organizational scaling into a single year. Different executives excel at distinct developmental stages of an enterprise. As a company scales rapidly, changes in personnel are a natural reflection of evolving organizational needs rather than an operational crisis.

The foundation focuses primarily on eradicating major human diseases, scientific discovery, and societal AI resilience. We have directed substantial funding toward research initiatives such as curing Alzheimer's disease. Within the next decade, advanced reasoning models will fundamentally transform biology and medicine, enabling breakthroughs that were previously unimaginable for human researchers working alone.

Halting technological progress out of fear would be a severe mistake that condemns humanity to continued suffering from incurable diseases. The historical trajectory of human civilization has always involved mastering powerful, double-edged discoveries. The responsible path forward is not retreating into the past, but actively developing defensive capabilities, monitoring infrastructure, and biological safeguards so we can unlock life-saving upsides safely.

We already pause training unilaterally whenever model capabilities advance faster than our alignment and verification techniques. At an industry level, companies will require clear regulatory mechanisms and antitrust exemptions to coordinate safety measures legally. Internationally, multilateral agreements between major powers such as the United States and China on shared safety standards and verifiable thresholds are essential to prevent dangerous capability race conditions.

We have invested immense engineering and research resources into ensuring conversational models do not encourage self-harm or substitute for genuine human connection. While AI has the potential to become a powerful mental health support tool, handling crisis situations requires immense care and collaboration with medical professionals. We actively consult psychiatric experts to refine model behaviors, provide crisis interventions, and direct vulnerable individuals toward human professional care.