AI Bros Are Having a Meltdown Because the Associated Press Stylebook Clarifies That AIs Don't Have Feelings

A short Futurism news piece about a post the AP Stylebook put on X on 23 September 2026. Its opening line says AI systems “do not think, feel, want or understand” and tells writers to avoid language that gives them human characteristics. Futurism quotes only that line. The post itself goes on to define anthropomorphizing and to point writers toward an alternative, and the article does not quote that part.

The article collects three objections, all from people now working at AI labs. Alan Rozenshtein says a stylebook should not make confident metaphysical pronouncements. Henry Shevlin points out that AP accepts anthropomorphic language about fictional robots in film reviews. Dean Ball says AIs clearly think and understand, and that a definition of “think” that excludes systems able to resolve a Millennium Problem is the flawed one. All three affiliations check out. Rozenshtein is on leave from the University of Minnesota Law School on Anthropic’s Rule of Law team, as the school announced in August. Shevlin became Google DeepMind’s philosopher in May and keeps a part-time Cambridge post. Ball joined OpenAI as head of strategic futures in July, according to the Foundation for American Innovation.

Landymore replies that there is no scientific consensus that AI systems have human-like emotions. He argues that caution in a newsroom is reasonable when an industry worth trillions of dollars profits from presenting its products as alive. He says the Millennium Problem claim behind Ball’s example is itself controversial. Quanta Magazine’s reporting on OpenAI’s Navier–Stokes result helps here: the proof was machine-checked in Lean, and the dispute is about credit, meaning how far it drew on outside mathematicians’ work (including unreleased work), not about whether it holds.