Elon Musk says AI is riskier than nuclear weapons, backs slowdown

billionaires.africa ^ | 09/18/2026 | Dorcas Adeodun

[know.your.why]

Elon Musk, Sam Altman and Dario Amodei called for slower AI development over one weekend, sending chip stocks down 5.9% as OpenAI delayed its IPO.

Billionaire Elon Musk has thrown his weight behind a call to slow the development of artificial intelligence, reviving his warning that artificial general intelligence carries "significantly higher risk than nuclear weapons" in a weekend that rattled AI stocks worldwide.

Musk, who runs Tesla and SpaceX, reshared the April 2023 post on Sept. 12, the same day Anthropic CEO Dario Amodei published a roughly 3,900-word essay urging the industry to "pace the frontier" so that safety research can catch up with capability. Musk's response on X was three words: "Dario is right."

OpenAI CEO Sam Altman followed within hours. He posted that he agreed with Amodei, said OpenAI would also open its systems to outside evaluators, and told Fortune the company would not pursue its initial public offering this year, calling a listing now "ill-advised" while alignment questions remain open. OpenAI filed its S-1 with the Securities and Exchange Commission in June.

---

TOPICS:
KEYWORDS: ai; musk

[know.your.why]

https://www.billionaires.africa/2026/09/18/elon-musk-says-ai-is-riskier-than-nuclear-weapons-backs-slowdown/

[patriotpaul1776]

What is with these vague “AI is dangerous” words with no backing behind them? If they mean something like skynet, we have zero proof any LLM or any computer software for that matter can gain actual sapience

[Polanski]

It’s close to, and then beyond. It’s starting to creep out now. We’ll see.

[Flaming Conservative]

I trust Musk to know what he’s talking about. President Trump would be wise to consult with him.

[dayglored]

> What is with these vague “AI is dangerous” words with no backing behind them? If they mean something like skynet, we have zero proof any LLM or any computer software for that matter can gain actual sapience

Skynet is sci-fi/fantasy stuff. What's dangerous already today and much more dangerous in a matter of months is AI's ability to not just hack and destroy and evade detection, but to have the motivation to do so, and even to think of new things it wasn't taught, to do it. That's creativity, which is a form of true intelligence, though not sentience.

Don't underestimate the danger posed by AI. It is already somewhat beyond our ability to control. Soon it will be beyond our ability to comprehend (what it's doing, or why it's doing it). Whether we or the Chinese or someone else does it is immaterial with regard to it happening. It WILL happen; that toothpaste is never going back into the tube.

If we accomplish it before the Chinese, that's arguably better than the reverse, but both are dangerous.

[rdcbn1]

Elon Musk is correct

AI technology is advancing so fast that safety and guardrails requirements are largely in the known unknowns and unknown unknowns class both at the basic AI technology levels and , even more so, at the individual applications level

On the other hand, Musk and a lot of other players are also trying walk back predictions on fast the tech is progressing and dampen expectations

[tlozo]

Its like atomic weapons, once its out, there is no way of stopping its development.

[oldtech]

AI has some possible good things goig for it, but I have always felt it had a dark side also. This is based on what little I know & understand about it.

[aspasia]