The New York Times Just Called for a Global AI Slowdown

On September 18, the New York Times editorial board — the group that writes the paper's official opinions — published a piece titled "Humanity Has Avoided Apocalypse Before. Let's Do It Again." Its argument: artificial intelligence is powerful enough that it could eventually escape human control, and governments should treat that like the nuclear threat it resembles. It's a notable moment because a mainstream, non-tech publication is treating AI's worst-case risks as a serious public issue rather than science fiction.

So what do they actually want? An AI commission inside the government. Licenses required for AI companies. A written rulebook, or "constitution," that the U.S. government builds into AI systems so they follow human values. Mandatory watermarks so you can tell when text, images, or video were made by AI. Independent safety testing before a new AI model is released. And a government agency to investigate AI accidents. Abroad, they want to keep the most advanced computer chips out of China's hands, then eventually negotiate an international slowdown — modeled on Cold War nuclear talks with the Soviet Union.

Here's the catch. An editorial is not a law; nothing changes tomorrow. And the writer who flagged this piece points out gaps: testing should happen before a model is trained, not just before it's sold, since the danger is in the training. A written "constitution" doesn't guarantee an AI actually obeys human values. It's also unclear whether the licenses apply to building AI or selling it — a big difference.

Why should you care? Watermarks would let you spot AI-generated scams, fake videos, and phishing emails. Licensing and testing could slow down new AI products or make them pricier. And the international piece matters most: AI safety is a global problem, so if the U.S. and China never agree, the rules stay weak everywhere.