Why Elon Musk Says We Should Slow Down AI
Elon Musk publicly backed Anthropic CEO Dario Amodei's call for top AI labs to slow down, posting "Dario is right" on Sept. 12, 2026. X
Amodei had asked artificial intelligence companies to deliberately slow work on frontier models, the most powerful systems. Anthropic pointed to very fast progress and the risk of recursive self-improvement, where AI helps make even smarter AI, like a printer that can print a better printer. The Wall Street Journal
Amodei said artificial intelligence will beat humans at most thinking tasks within a "small number of years." Reuters
Why that speed matters for your money is simple. Big payoffs and job disruptions would land in the next few years, inside normal business spending plans, not decades away.
Amodei links that speed to compute, meaning raw computing power. In prepared remarks for the AI Safety Summit, he said rapid progress is driven by available computation growing by 8x per year and unlikely to slow down. Anthropic
The plain takeaway from that number is that computing power sets the speed limit. Skilled people and smart ideas matter, but having enough computers decides who moves fastest.
Amodei leads Anthropic. Anthropic The company has published his statement titled 'Dario Amodei on American AI leadership' and, on July 27, 2026, his post titled 'Our position on open-weights models.' It has also published 'Policy on the AI Exponential' sharing two policy proposals to prepare for AI progress. Musk was named to TIME's 100 Most Influential People in AI for 2026. TIME
The bigger problem here is getting rivals to act together. Each lab gets some safety and reputation benefit from holding back but pays the full price in lost capabilities, partnerships and idle computers. Backing from a rival makes it safer to speak up. It does not enforce anything.
In my view, savers should separate the idea from how likely it is. The idea is to slow top-level training and use until safety rules catch up. How likely that is depends on clear promises, shared safety tests and punishment for cheating. The verified statements do not include those steps. Without them, markets expect big companies to keep spending their cash to build more.
For ordinary investors, there is a money clash to understand. If usable computing power keeps rising 8-fold a year, leaving computers idle is costly. That pushes firms to use them. Strict limits on sharing powerful models help big firms with closed products and business customers. Loose sharing cuts their edge but spreads use. Stock and bond investors will read the same safety talk as possible new rules, good or bad depending on time horizon and business type.
