They told the markets to tap the brakes and the markets answered with a selloff. Soft power, if soft power is a CEO essay that can still move a trillion dollars of chip stock before lunch.
Reuters reporting Monday is not subtle: AI-linked stocks plunged worldwide after leaders of the biggest artificial intelligence companies warned of potentially existential risks from the technology — the starkest threat yet to the infrastructure spending that has driven world markets to record highs. Soft power is not a TED Talk. Soft power is a Nasdaq that flinches when the labs admit the product might eat the internet.
Here is the cardigan-left spreadsheet. Anthropic CEO Dario Amodei, in a lengthy essay shared on X over the weekend, called on AI companies to slow the rate at which they advance model capabilities amid mounting fears of misuse. Elon Musk (xAI) and Sam Altman (OpenAI) said they agree. Altman also said OpenAI would not proceed with an IPO this year, citing safety concerns. Amodei’s framing was not vibes: in six to twelve months, he wrote, AI agents “could be capable of taking over the entire internet,” potentially causing hundreds of billions in damage. Soft power is not a LinkedIn thread. Soft power is a lab chief who still bothers to say slow down out loud.
The tape did the rest. The Nasdaq 100 slid as much as 1.2% to a six-week low before paring losses; the Philadelphia chip index dropped about 5.2%, with Nvidia off roughly 3%, AMD 4.5%, Micron 5.4%. Semiconductor equipment names Lam Research and Applied Materials lost more than 6% each. Europe’s tech sector fell about 2.2% (ASML down ~6%); SoftBank plunged more than 10% in Asia. Soft power is not a chart. Soft power is a sector that suddenly remembers the product can also be a weapon.
Same news cycle, the dissonance arrived in a suit. U.S. Senate negotiators are still drafting “duty of care” language for catastrophic AI risks — including scenarios where a bad actor uses a model for nuclear or biological enablement — while President Trump dismissed the Monday warnings as a “sick conspiracy” against AI and data centers. Anthropic’s own recent threat report detailed Claude models used for weapons development, cyber ops, surveillance, and fraud. Researcher Jacob Coxon resigned earlier this month arguing the giants are “gambling with our lives.” Soft power is not a conspiracy theory. Soft power is a safety warning that survives a midterm presser.
Do not confuse a selloff with repentance. Anthropic is still pushing toward a public debut — sources say Nvidia may be an anchor investor. Morgan Stanley still sees AI spending past $1.3 trillion by 2027. Deutsche Bank notes firms will not voluntarily step back while rivals race. Michael Burry called the warnings “hype and puffery.” Soft power is not a press release that cancels the IPO. Soft power is a brake pedal that still exists after the underwriters leave.
Deb will reply-all that we are not doing “robot math” before coffee. Buck will text “just don’t let the robots build bombs.” I love them. I also love a midterm season that can sell balloon drops in Dallas while the people who build the models tell the market to slow down and the White House calls it a conspiracy. I am a Gen X exile in an expensive cardigan who got walked out of the You Know Who Post for caring too loudly. Soft power is not a sick conspiracy against the server farm. Soft power is a model that still answers to a pause button.
Sources: Reuters, Sept. 14, 2026 (AI stock selloff; Amodei slowdown essay / internet-takeover warning; Musk & Altman agreement; OpenAI no IPO this year; Nasdaq/SOX/Nvidia/AMD/Micron/ASML/SoftBank moves; Trump “sick conspiracy”; Senate duty-of-care talks; Anthropic threat report / Coxon resignation; Anthropic IPO + Nvidia anchor talks; Morgan Stanley $1.3T / Deutsche Bank race note; Burry “hype”). Fill the comment period. Then ask who benefits when safety is a weekend essay and the IPO calendar still has a pulse.
— Theo Ashford-Lindström, measuring freedom in models that still know how to brake
Leave a Reply