Should You Be Concerned About AI?
I’ve been building apps with AI since January 2023, starting with ChatGPT 3.5. The technology amazed me then. Today’s models are so much more powerful that I can only imagine what’s coming next.
In August of 2025, our COO, Mark Douglas, shared an article called Situational Awareness by Leopold Aschenbrenner. I think every business leader should read it. Fair warning: it’s long. You can also download the full 165-page report as a PDF if that’s easier.
Here’s why I’m sharing this. I’m worried. Not just for myself, but for every person who should be paying attention to how AI gets adopted and what it means.
Two things keep me up at night.
The first is bad actors. Countries like China, Russia, North Korea, and Iran are using AI with bad intentions. We hear new stories nearly every day. DeepSeek, China’s AI tech, already has the stock market on edge. We need tighter security now, or we risk an awful future for humanity.
The second is small business owners. I have friends building AI apps on platforms like Replit, Emergent, and Cursor. I’m worried for them. The market is filling up fast with AI apps priced at almost nothing. If a giant like Amazon, Google, or Microsoft decides to compete, smaller players get crushed. These big companies are now building their own power plants. That kind of advantage is hard to beat.
I’ve been burned by technology before.
In 1989, I started Elledge Photographic Lab. We made high-end slide duplicates for the world’s largest stock agencies. Our work hung in airports as displays and murals. Then digital cameras arrived. The first ones were rough, but they caught up to film fast. Large format printing followed. There was no need for my business anymore.
Don’t get me wrong. I’m not all doom and gloom. AI will do real good, especially in healthcare. I just want you to be careful.
This paper came out in June 2024. By today’s standards, that’s old news. But it’s worth a look. The predictions are now testable against what actually happened.
Things are moving fast.
Below is a summary of the paper. You can read the full report here.
Executive Summary: Situational Awareness
Leopold Aschenbrenner is a former OpenAI researcher who worked on the Superalignment team. In June 2024, he published a 165-page essay called Situational Awareness: The Decade Ahead. His thesis is direct. AGI is coming this decade. Most people, including most business leaders, are not ready.
Here’s what he argues, in plain English.
1. AGI by 2027 is plausible.
Aschenbrenner traces the path from GPT-2 to GPT-4. In four years, AI went from barely stringing together sentences to passing college exams. He expects another jump that big within four years. The math is simple. Compute keeps growing. Algorithms keep improving. Tools keep getting better. Add it up and AI starts doing the work of AI researchers themselves. That puts AGI on the table by 2027.
2. Superintelligence comes next.
Once AI can do AI research, it speeds up its own progress. Aschenbrenner calls this the intelligence explosion. Picture millions of AI systems working day and night, faster than any human, on building smarter AI. A decade of progress could happen in one year. The result is superintelligence. AI that is smarter than every human alive, by a wide margin.
3. The build-out is enormous.
Companies are racing to build trillion-dollar compute clusters. Power, data centers, and chips will need massive investment. American electricity production will grow by tens of percent. This is industrial mobilization at a scale we have not seen in 50 years.
4. Security at AI labs is weak.
This is one of Aschenbrenner’s loudest warnings. Top AI labs treat security like a SaaS startup. Foreign intelligence services treat them like soft targets. He argues China can already steal the key research. If the U.S. does not lock down lab security, the CCP gets superintelligence too.
5. We do not know how to control superhuman AI.
This is the alignment problem. Today’s methods for keeping AI on task work for current models. They will not scale to systems much smarter than us. Aschenbrenner is not a doomer, but he is worried. The jump from human-level AI to superintelligence may happen in less than a year. There is no time to figure it out as we go.
6. The race against China is real.
Whoever gets superintelligence first gets a decisive economic and military edge. Aschenbrenner argues America must lead. A U.S. lead also buys time to handle safety carefully.
7. The government will take over.
He predicts that by 2027 or 2028, the U.S. government will run the AGI program. He calls it “The Project.” Private startups, in his view, cannot handle a national security project of this size.
The bottom line for executives.
Aschenbrenner closes with a framework he calls AGI Realism. Three points:
- Superintelligence is a national security issue, not just a tech story.
- America must lead the race.
- We cannot afford to screw it up.
You may not agree with every prediction. Many smart people do not. But the paper is worth your time because it lays out a specific scenario with specific timelines. Now, almost two years later, you can compare what he predicted with what actually happened.
That is a useful exercise for any leader trying to plan around AI.