Altman, Amodei, and Musk publicly warn that artificial general intelligence could end civilization. Then they keep building anyway. The real reason is darker than any headline has admitted.

By Kristal Thapa | Technology | 10 min read

Imagine a surgeon standing over an open patient, holding a scalpel, saying this procedure might kill the patient, and then continuing anyway.

That is essentially what the most powerful people in Silicon Valley are doing right now. They built the tools. They raised the alarm. And they are still building.

These are their own words, on the record, in public. Not conspiracy theories. Direct statements from the people writing the code.

The CEOs Who Said AGI Is Dangerous

The warnings on artificial general intelligence risk did not come from activists or fringe forums. They came directly from the leadership of the world's most advanced artificial intelligence companies.

In May 2023, the chief executives of OpenAI, DeepMind, and Anthropic signed a joint statement alongside hundreds of prominent AI researchers. The language was brief and unflinching.

Mitigating the risk of extinction from AI should be a global priority alongside pandemics and nuclear war.

Not job losses. Not privacy erosion. Extinction. The end of humanity as a biological species. These were not fringe voices. They were the people running the laboratories responsible for the most advanced AI systems ever built.

Elon Musk, who co-founded OpenAI before departing to launch xAI, has placed his personal estimate of the probability that AI causes human annihilation at around 20 percent. Dario Amodei of Anthropic has publicly cited his own probability of catastrophic AI outcomes at roughly 25 percent.

These figures come from people who control the most consequential AI research organizations on the planet. One in four chance that their own technology destroys civilization. And the funding rounds keep growing. The data centers keep expanding. The artificial general intelligence timelines keep compressing.

The question is not whether they understand the risks. The question is why they continue regardless.

Key figures telling us to be afraid:

• Sam Altman, CEO of OpenAI, has repeatedly described AGI as potentially the most transformative and dangerous technology in human history
• Dario Amodei, CEO of Anthropic, assigns a 25 percent probability to catastrophic AI outcomes
• Elon Musk, founder of xAI, estimates a 20 percent chance of AI-caused civilizational collapse
• Geoffrey Hinton, the godfather of deep learning, resigned from Google in 2023 specifically to speak freely about existential AI risk.

Why They Keep Building It Anyway

This is where the psychology becomes genuinely fascinating, and honestly, a little dark.

Anthropic CEO Dario Amodei offered one of the most candid admissions in recent technology history. Almost every decision I make is balanced on the edge of a knife, he explained, noting that building too fast risks losing human oversight of artificial general intelligence, while building too slowly risks authoritarian states reaching it first and shaping it to their own political ends.

That is the trap. That is the logic running Silicon Valley at this precise moment.

These executives are not reckless. Most appear genuinely worried. But they believe that stepping aside and allowing someone else to reach AGI first is more dangerous than pressing ahead with safety measures built in from the start.

Researchers at MIT Technology Review drew a direct historical parallel. This is Silicon Valley's Oppenheimer moment. The physicists who built the atomic bomb understood exactly what they were creating. They built it anyway, partly because they feared Nazi Germany would reach it first. The underlying logic has not changed much since 1945. Only the acronym has.

That fear is rational given real geopolitical dynamics. Emerging technology races, from quantum computing to large language models, increasingly define national power. The pressure inside boardrooms and research divisions is not imaginary. It shapes every hiring decision, every product roadmap, and every infrastructure investment made in the sector.

This is the arms race mentality applied to the most consequential technology in recorded history. And it is driven not by recklessness, but by fear itself.

The Financial Clause Nobody Talks About

Now for the part that most mainstream outlets consistently underreport.

OpenAI has a commercial partnership agreement with Microsoft, which has committed over 13 billion dollars in investment to the company. Inside that agreement sits a clause that has received minimal press scrutiny.

According to TIME, the agreement reportedly ties the definition of AGI achievement to a profit milestone for early investors, with OpenAI's board retaining final authority over when that threshold is reached.

That is a financial trigger, not a philosophical one. It changes everything about how we should interpret the AGI timeline debate coming from OpenAI's leadership. The goalpost is not a scientific consensus. It is investor returns.

When Sam Altman stated in a Forbes interview in February 2026 that we basically have built AGI, or very close to it, Microsoft CEO Satya Nadella pushed back publicly. I do not think we are anywhere close to AGI, Nadella said. Altman then stepped back, acknowledging that reaching AGI genuinely would still require many significant breakthroughs across multiple research domains.

This exchange reveals something important. The definition of AGI is fluid, contested, and tied to commercial interests. We are building the most consequential technology in human history toward a goalpost that the people building it cannot agree on.

That opacity carries real consequences, particularly for workers and industries already experiencing AI-driven disruption at scale.

The Race Logic: If Not Us, Then Who?

To understand why even frightened people continue building, you need to understand one belief that sits at the foundation of Silicon Valley thinking: artificial general intelligence is going to happen regardless of who decides to slow down.

Altman wrote on his personal blog in January 2025 that they are now confident they know how to build AGI as they have traditionally understood it. In their worldview, the train has already left the station.

If OpenAI stops, Google DeepMind continues. If the United States decelerates, China accelerates. If established laboratories pause, dozens of smaller teams across the world do not. The belief, right or wrong, is that AGI will exist eventually, and the only meaningful question is who controls it when it arrives.

Altman's published roadmap makes this ambition concrete:

• 2025: arrival of AI agents capable of performing real cognitive work
• 2026: systems that produce novel scientific insights independently
• 2027: physical robots operating with real-world independence
• Early 2030s: AI-powered individuals doing the work of entire specialist teams

The geopolitical stakes reinforce this logic at every level. Enormous capital poured into AI chip infrastructure, and the integration of autonomous AI systems into military strategy adds layers of urgency that no chief executive can rationally set aside. When advanced AI systems could potentially shift the global balance of power within a single technology cycle, the incentive to pause is structurally weak.

There is an uncomfortable truth buried here. Racing to control something dangerous because you fear others misusing it is still racing. The fear does not slow the development. It simply changes who sits at the wheel.

What the Data Actually Shows

Strip away the rhetoric and look at what peer-reviewed research and credible surveys actually reveal about existential risk from advanced artificial intelligence.

A 2022 survey of AI researchers found that the majority assigned a 10 percent or greater probability to AI causing existential catastrophe through humanity's failure to maintain meaningful control over advanced systems. A separate large-scale survey across more than 2,700 AI researchers estimated a 10 percent chance that AI will outperform humans across most cognitive tasks by 2027.

A 10 percent probability of civilizational-scale transformation within two years is not a footnote. It is a fire alarm.

The AI Safety Clock, launched by the International Institute for Management Development in September 2024, began its count at 29 minutes to midnight. By February 2025, it moved to 24 minutes. By March 2026, it stood at 18 minutes. Eleven minutes closer to the perceived threshold of irreversible crisis in 18 months of recorded progress.

The Financial Reality Behind the Ambition

OpenAI's own financial position tells a revealing story about the gap between idealism and commercial pressure. The company has reported substantial annual losses and serves more than 800 million weekly active users as of late 2025, according to its own disclosures, with profitability not projected until 2029.

You do not sustain losses of that scale building a product you believe will fail. You do it when you believe the potential upside is so transformative that conventional financial logic no longer applies. That conviction, even when accompanied by genuine fear, is what keeps the laboratories running around the clock.

Anthropic has been raising capital at valuations reported in the hundreds of billions of dollars, while its CEO simultaneously publishes warnings about existential risk from the very products his company ships. Fear and ambition occupy the same boardroom. That is not a contradiction the market seems troubled by.

The Contradiction at the Core of Big Tech

MIT Technology Review surfaced an observation that deserves far more attention than it received.

Katja Grace, lead researcher at AI Impacts, observed in October 2025 that these CEOs say artificial superintelligence will kill us, but they are laughing while they say it. That detail reveals a specific psychological posture.

The people building AGI have concluded that the danger is real, that the odds still favor a positive outcome, and that the risk of a competitor reaching it first outweighs the risk of continuing themselves. They laugh not from cruelty, but because the situation is genuinely absurd and they know it.

Dario Amodei published a lengthy essay in early 2026 arguing that humanity is about to receive enormous technological power and that our social and political institutions may lack the maturity to wield it responsibly. He wrote this essay while leading the company that builds the technology he warns about, and he acknowledged within the essay that the intervention might ultimately be futile.

That is not hypocrisy for its own sake. That is what happens when you believe you cannot stop something but still believe you might influence how it lands.

The Investor Narrative Nobody Examines Closely Enough

There is a financial dimension to this contradiction that receives insufficient scrutiny. Anthropic closed significant funding rounds at elevated valuations during the same period that Amodei published his most prominent existential risk warnings. Altman faces ongoing pressure to present investors with evidence that transformative breakthroughs remain imminent and fundable.

This does not mean the risks are fabricated. It does mean that the narrative surrounding artificial general intelligence serves multiple simultaneous purposes: genuine scientific alarm, competitive positioning between laboratories, investor confidence management, and geopolitical signaling. All from the same executives. All in the same press cycle.

Understanding that layering of motivations is essential to reading any AGI-related announcement critically. Even the commercial pricing structures of current AI products reveal an industry betting everything on a future that remains unproven at scale.

What Happens Next: The Road Ahead for Artificial General Intelligence

The most honest answer to where this leads is that nobody fully knows. And that itself is extraordinary.

We are at a point in history where the people building the most powerful technology ever attempted openly admit they do not fully control its trajectory. What separates artificial general intelligence from every previous technology is this: most technologies amplify human capability in one specific domain. A faster aircraft covers more distance. A better communication network connects more people.

AGI, by definition, amplifies everything simultaneously, including the human capacity for miscalculation, conflict, and harm.

The global regulatory landscape remains fragmented. Hardware progress continues regardless of what policymakers decide. Security researchers already warn that AI's expanding autonomy creates attack surfaces that current cybersecurity frameworks were never designed to address.

In 2025, a Future of Life Institute open letter signed by five Nobel Prize laureates called for a prohibition on superintelligence development until a broad scientific consensus exists that it can be built safely. That letter received a week of coverage. Then the laboratories kept building.

The most alarming aspect of the AGI story is not the technology itself. It is the logic trap that intelligent, well-informed, genuinely worried people cannot seem to escape.

If we stop, someone worse takes over. If we move too fast, we lose control. There is no clean exit. There is only the question of which risk to accept, and who gets to make that choice on behalf of everyone else on this planet.

Meanwhile, AI tools already replace entire categories of traditional software, major technology companies quietly restructure entire divisions in anticipation of the AGI era, and society grapples with the effects of algorithmic AI on younger generations, all while the existential debate unfolds above most people's everyday awareness.

The real question is not whether AGI is coming. Based on every credible indicator, some version of it is. The real question is who shapes the governance frameworks of the world it creates, and whether those frameworks will exist before they are needed.

At this moment, the answer is a small group of chief executives who, by their own frank admission, are frightened, financially motivated, and still at the wheel. The investors are placing bets worth hundreds of billions. The regulators are scrambling to keep pace. The scientists are signing letters they know will be ignored.

And the rest of us are watching one of history's most consequential experiments unfold in real time.

Should the people building AGI also be the ones who decide when it is safe? Drop your perspective in the comments.