# Anthropic Researcher's Extinction Warning Revives the Push for United States AI Rules

President Trump said he is not worried that artificial intelligence will wipe out humanity, as Europe debates whether data, not models, is where it can still compete.

- Published: 2026-09-11T05:17:40.652Z
- Canonical: https://polylog.news/2026-09-11/anthropic-researcher-s-extinction-warning-revives-the-push-f
- Publisher: Polylog (Global desk)
- Section: tech
- Sources: [The Japan Times](https://www.japantimes.co.jp/business/2026/09/11/us-ai-rules-anthropic-safety/), [RIA Novosti](https://ria.ru/20260911/tramp-2116965767.html), [Financial Times](https://www.ft.com/content/f054f927-b512-452a-b494-ea53f5ac1079)

Calls for federal rules on artificial intelligence intensified in Washington this week. [The Japan Times reported that alarm grew after an Anthropic researcher said the people building artificial intelligence "earnestly believe that it could kill us all by the end of the decade"](https://www.japantimes.co.jp/business/2026/09/11/us-ai-rules-anthropic-safety/), a statement from inside one of the leading laboratories rather than from an outside critic.

The White House has not changed course. [RIA Novosti reported that President Donald Trump said he is not worried that artificial intelligence will lead to the extinction of humanity](https://ria.ru/20260911/tramp-2116965767.html). The administration has treated speed of deployment, not restraint, as the American advantage.

Europe is arguing about a different question. [The Financial Times argued that data is the one area where Europe can still be sovereign and holds the greatest potential to generate growth](https://www.ft.com/content/f054f927-b512-452a-b494-ea53f5ac1079), an implicit concession that the contest over frontier models and the chips that train them has already been settled elsewhere.

Three jurisdictions are therefore heading in three directions: American acceleration with contested safety rules, European regulation built around data rights, and Chinese construction of a domestic hardware and model stack under export controls. Companies deploying these systems across borders will carry the cost of all three regimes at once.

## What this means

Regulatory divergence is a margin problem before it is a safety problem, because a company selling the same model in the United States, the European Union and China must build separate compliance, data handling and audit processes for each. That favours the largest platforms, which can absorb fixed compliance costs, and it disadvantages smaller developers and the enterprises that buy from them. For investors in artificial-intelligence infrastructure, a binding American safety rule would extend deployment timelines, which is the variable that determines when the current capital spending converts into revenue.

## What to watch

- Whether any federal artificial-intelligence bill reaches a floor vote in Congress, the difference between rhetorical alarm and an actual compliance cost.
- Whether more researchers inside leading laboratories make public safety statements, because internal dissent moves legislators faster than outside advocacy does.
- How Europe defines data sovereignty in practice, since rules on training-data access determine whether European firms can build competitive models at all.