Sam Altman posts after a Molotov cocktail was thrown at his home: I love my family, and I believe AI belongs to everyone.

Author: Sam Altman

Translation: Deep Tide TechFlow

Deep Tide Guide: Someone threw a Molotov cocktail at Sam Altman’s house at 3:45 a.m. He rarely shares a family photo publicly, hoping it might deter the next person from throwing something at his home. This article is not just an account of the attack; he also fully articulates his beliefs about AI for the first time: AI must be democratized, and a few labs should not decide humanity’s future, while the temptation of “once you see AGI, you can’t go back” fills this field with Shakespearean dramatic conflicts.

This is a photo of my family. I love them more than anything.

I hope images have power. Usually, we keep a fair amount of privacy, but in this case, I’m sharing a photo in the hope it might discourage the next person from throwing a Molotov cocktail at our home, regardless of how they view me.

The first person did this last night, at 3:45 a.m. Fortunately, it bounced off the house, and no one was hurt.

Words also have power. A few days ago, there was a provocative article about me. Yesterday, someone told me they believed that article appeared at a time when people are extremely anxious about AI, making my situation more dangerous. I didn’t think much of it at the time.

Now I am awake in the middle of the night, angry, and I realize I underestimated the power of words and storytelling. It seems like a good time to talk about a few things.

First, my beliefs.

Working toward prosperity for everyone, empowering all people, advancing science and technology—these are moral obligations for me.

AI will be the most powerful tool to extend human abilities and potential. The demand for this tool is essentially limitless; people will use it to do incredible things. The world should have abundant AI, and we must figure out how to make that happen.

Not everything will go smoothly. Fear and anxiety about AI are reasonable; we are witnessing, perhaps for the first time in history, one of the greatest societal shifts ever. We must get safety right; this is not just about aligning a model—we urgently need society-wide responses to new threats. This includes new policies to help navigate difficult economic transitions toward a better future.

AI must be democratized; power cannot be overly concentrated. Control over the future belongs to everyone and their institutions. AI needs to empower individuals; we need collective decisions about our future and new rules. I don’t believe it’s right for just a few AI labs to make the most important decisions about our future form.

Adaptability is crucial. We are all learning new things at a very rapid pace; some of our beliefs will be correct, some will be wrong, and sometimes we need to change our minds quickly as technology develops and society evolves. No one fully understands the impact of superintelligence, but it will be enormous.

Second, some personal reflections.

Looking back on my first decade at OpenAI, I can point to many things I am proud of and a bunch of mistakes.

I am thinking about our upcoming trial with Elon, and I remember how strongly I resisted agreeing to his desire for unilateral control over OpenAI. I am proud of that, proud of the narrow path we took at the time that allowed OpenAI to continue existing, and of all the achievements that followed.

I am proud that I did not shy away from conflict, even though it brought me and OpenAI great pain. I am not proud of how I handled conflicts with the former board, which caused significant chaos within the company. I made many other mistakes along OpenAI’s wild trajectory; I am a flawed person, caught in an extraordinarily complex situation, trying to improve every year, always working for the mission. From the beginning, we knew how risky AI was, and that personal disagreements among well-meaning people would be greatly amplified. But experiencing these intense conflicts firsthand and often having to arbitrate them is another matter, and it came at a heavy cost. I apologize to those I hurt and hope I can learn faster.

I am also very aware that OpenAI is now a major platform, not a small startup; we now need to operate in a more predictable manner. The past few years have been extremely tense, chaotic, and high-pressure.

But what I am most proud of is that we are fulfilling our mission, which initially seemed almost impossible. Overcoming all obstacles, we figured out how to build very powerful AI, how to raise enough capital to develop the infrastructure to deliver it, how to create product companies and businesses, how to deliver large-scale, relatively safe, and robust services, and more. Many companies say they want to change the world; we truly have.

Third, some thoughts on this industry.

My insights from the past few years, and my view on why there is so much Shakespearean drama among companies in our field, boil down to this: “Once you see AGI, you can’t unsee it.” It has a real “Ring of Power” dynamic that drives people to do crazy things. I’m not saying AGI itself is the ring, but rather the totalitarian philosophy of “becoming the one who controls AGI.”

The only solution I can think of is moving toward broad sharing of technology, with no one holding the ring. Two obvious ways to do this are empowering individuals and ensuring democratic systems retain control.

It’s crucial that democratic processes remain stronger than corporations. Laws and regulations will change, but we must work within democratic processes—even if they are messy and slower than we’d like. We want to be a voice and stakeholder, but not hold all the power.

Much criticism of our industry comes from sincere concern about the high risks of this technology. That’s very reasonable, and we welcome good-faith criticism and debate. I empathize with anti-technology sentiments; obviously, technology isn’t always good for everyone. But overall, I believe technological progress can make the future incredibly better—for your family and mine.

When we have that debate, we should tone down rhetoric and strategy, and try to have fewer explosions at home—metaphorically and literally.