For about a week, it looked like the AI industry had reached a rare moment of self-restraint. After Anthropic researcher Jacob Coxon resigned, warning in a post now seen more than 170 million times that his former employers were "racing straight to self-improving superintelligence and gambling with our lives," the leaders themselves seemed to agree. Dario Amodei published an essay urging companies to "slow the pace." Sam Altman, Elon Musk and Google DeepMind's Demis Hassabis all signalled support; Altman went further and shelved OpenAI's IPO for the year. It was the reaction almost nobody expected. This week the world outside the labs delivered its verdict, and the neat consensus began to come apart.
Wall Street answered first, and bluntly. On Monday AI-linked stocks sold off worldwide, with the Nasdaq 100 sliding 1.7 percent and the Philadelphia chip index dropping 6 percent. Nvidia fell 3.5 percent, Micron nearly 7, and equipment makers Lam Research and Applied Materials shed 8 and 7 percent. SoftBank plunged as much as 13 percent in Tokyo. The message from investors was awkward for an industry that had just discovered caution: markets have been "running hot based on AI spending," as Interactive Brokers' Steve Sosnick put it, and any real slowdown threatens the earnings story propping up the whole rally.
Then came the president. Donald Trump dismissed the safety warnings as a "SICK conspiracy" and a "hoax," casting the people calling for guardrails as "Conspiracy Theorists, Treasonists, Traitors, and Leakers," and arguing that any brake on the US data-center buildout would only hand the advantage to China. It was less a policy than a political line in the sand, and House Speaker Mike Johnson reinforced the practical reality: AI legislation "can't be rushed," and Congress has almost no calendar left before the midterms. A pile of bills exists, from a Frontier Act licensing regime to a proposed ban on "superintelligence," but none is close to becoming law.
Most revealing was the fracture inside the industry itself. Mark Zuckerberg said flatly that labs do not need to slow down, arguing that "alignment," not pace, will decide who wins, and that liability and competition already push firms to build safely. Yann LeCun, one of AI's so-called godfathers, was harsher, reminding everyone that Amodei once called GPT-2 too dangerous to release in 2019: "I made fun of them then. Everyone should make fun of them now." The united front that formed around Coxon's warning turned out to be a week old and already splitting along the industry's oldest fault line, between those who see existential risk and those who see a moat.
What is left is a debate that has escaped the people who started it. King Charles is due to host executives from Nvidia, OpenAI, Anthropic and DeepMind to discuss shared principles; Germany calls a halt "not viable," while Spain floats a nuclear-style non-proliferation pact. The lesson of the week is not that the slowdown failed, but that a voluntary pause was never the labs' to grant. The moment they asked for one, the markets that fund them, the president who wants to beat China, and the rivals who never agreed all pulled in different directions. Restraint, it turns out, is not a decision a handful of chief executives can make on a Saturday. It is a fight, and it has only just started.