AI CEOs Have Been Warning About AI Dangers for Years. Here Is What They Actually Did.
From Elon Musk's 2017 governors speech to last week's chorus of caution, the tech industry's most powerful voices have repeatedly called for AI regulation, while building the very systems they say scare them.
Key points
- A March 2023 open letter signed by Musk called for a pause on large AI experiments; two weeks later he announced his own new AI company.
- OpenAI CEO Sam Altman testified before the US Senate on 16 May 2023, asking Congress to create a new agency to regulate AI.
- A 22-word statement signed by Altman and then-Google DeepMind CEO Demis Hassabis on 30 May 2023 called AI extinction risk a global priority.
- The pattern of calls for regulation stretches back to computing pioneer Alan Turing, who warned in a 1951 lecture that AI machines would eventually "take control."
The names read like a tech industry roll call: OpenAI's Sam Altman, Anthropic's Dario Amodei, Google DeepMind's Demis Hassabis, Microsoft's Satya Nadella, Elon Musk. All, in recent days, said publicly that AI needs to slow down before it slips out of human control. Profit handsomely from it moving fast? Also yes.
The scepticism writes itself. But the calls are not new, and tracing them back reveals a pattern worth understanding.
How long has this been going on?
Longer than most people realise. Author Samuel Butler warned in 1863 that self-replicating machines could replace humans; his 1872 novel Erewhon depicts a society that banned machines entirely to survive. Alan Turing, the mathematician who laid the theoretical foundations for modern computing, said in a 1951 lecture that AI would "take control" once it surpassed human capability.
In 2000, Sun Microsystems co-founder Bill Joy argued that self-replicating robots posed a greater threat than nuclear weapons, partly because private companies, not governments, were building them.
The modern version, where the people funding AI ask others to regulate it, began with Musk.
What did the tech executives actually say, and when?
| Date | Who | What they called for |
|---|---|---|
| July 2017 | Elon Musk | Immediate government regulation of AI |
| December 2018 | Microsoft's Brad Smith | Laws governing facial-recognition AI in 2019 |
| January 2020 | Alphabet's Sundar Pichai | Regulation is "too important not to" pursue |
| May 2023 | OpenAI's Sam Altman | New federal agency to oversee AI models |
| May 2023 | Anthropic's Dario Amodei | AI could help create bioweapons; Congress must act |
| May 2023 | Altman, Hassabis and others | AI extinction risk equals pandemics and nuclear war |
Musk's 2017 speech to the National Governors Association is the clearest early example. He had already spent $38 million on OpenAI, already called AI "potentially more dangerous than nukes," and told the governors: "By the time we are reactive in AI regulation, it's too late."
In March 2023, he co-signed a letter from the nonprofit Future of Life Institute calling for a six-month pause on large AI experiments, the kind of large AI model, such as the systems behind ChatGPT, that can write or reason from prompts. Two weeks after signing, he launched his own AI company, xAI.
Our story the day before this one, "AI's Top Chiefs All Want a Slowdown. Should We Believe Them?", found the same rare unity among those same leaders, and noted that the reasoning behind it is more complicated than it looks.
Should ordinary people take these warnings seriously?
The warnings deserve attention, even if the messengers have complicated motives. When executives ask for regulation, they often shape what that regulation looks like, which can mean rules that favour large incumbents and make life harder for smaller competitors. That's worth keeping in mind.
What matters for anyone who uses AI tools at work or at home is simpler: these systems are changing fast, the people building them publicly admit they don't fully understand the risks, and no government has yet enacted broad rules. The warnings are real. The follow-through has been thin. The gap between the two is where to keep your attention.
Common questions
Does asking for regulation mean a company thinks its products are actually dangerous?
Not necessarily. Calls for regulation can reflect genuine concern or competitive strategy, sometimes both at once. A company that helps write the rules can benefit from them, which is why the substance of any proposed rules matters more than the fact that someone asked for them.
What has actually been regulated so far?
As of mid-2025, no broad federal AI law exists in the United States. The European Union passed the EU AI Act in 2024, the first wide-ranging AI law from a major economy, but its full enforcement phases in gradually over several years.
