What exactly Anthropic is proposing
The call appeared in a paper titled "When AI Builds Itself," which the company published on June 4. Its authors are Marina Favaro and Anthropic co-founder Jack Clark. In it, the company writes that it would be good for the world to have "the ability to slow down or temporarily pause the development of frontier AI" so that societal structures and safety research can keep pace with the technology. Anthropic ↗ Der Spiegel ↗ Tagesspiegel ↗In an interview for BBC's Newsnight, Jack Clark summed it up figuratively: according to him, artificial intelligence needs a "brake pedal," because it could reach a stage where it develops "without human input." BBC ↗
A machine that improves itself
At the heart of the concern is so-called recursive self-improvement — the notion of a system able to contribute to the design and improvement of subsequent generations of AI without substantial human help. Anthropic states that more than 80 percent of the code added to its production systems by May 2026 was written by its own model, Claude. Anthropic ↗ Le Figaro turned this into a headline: "The human role shrinks at every step of development." Le Figaro ↗According to the company, the goal of the pause is to give humanity time to adapt before the pace becomes uncontrollable. France 24 ↗
Warning while growing
It is precisely the timing that raises caution among European newsrooms. In the same week that Anthropic warns of losing control, media outlets describe its race with OpenAI to go public as a contest that goes beyond the markets themselves. Courrier International ↗ Austria's Der Standard comments on the whole rush with a terse headline: "Where this madness leads, we will soon find out." Der Standard ↗At the same time, the company is expanding the commercial and government deployment of its most powerful model, Mythos. According to the Financial Times, it has placed its engineers directly inside the US National Security Agency (NSA) to help deploy Mythos in offensive cyber operations Financial Times ↗, while the Belgian press reports on its deployment in the European financial infrastructure Euroclear. De Standaard ↗ The market mood is also illustrated by a curiosity from New York, where a luxury apartment in Brooklyn is being offered for sale in exchange for Anthropic shares. Naftemporiki ↗
How Europe reads it
Le Monde frames the proposal directly as a "coordinated global moratorium" and notes that it comes from a company that itself belongs among those the pause would be meant to restrain. Le Monde ↗ The call, meanwhile, lands in an environment where resistance to artificial intelligence is growing regardless. Der Standard notes that opposition to it is becoming both broader and more radical. Der Standard ↗ For the European Union, which is still only rolling out its AI Act in practice, a global moratorium is a goal a level more ambitious than what the EU's current tools cover.Safety, or slowing down the competition?
Anthropic is not the first in the industry to call for a brake, and criticism came quickly. Part of the industry, along with critics close to the Trump administration — including its informal adviser David Sacks — argue that the company is exaggerating worst-case scenarios and that the call actually serves as a way to slow down competitors under the guise of safety. SiliconANGLE ↗Anthropic itself concedes that a real moratorium would require major companies in multiple countries — chiefly the US and China — to stop simultaneously, under verifiable rules. It likens the problem to nuclear arms control treaties, but adds that with AI it is even harder: training a model can be hidden more easily than a missile silo, and the temptation to quietly continue would be enormous.
What hotinfo is watching
- Whether OpenAI and Google DeepMind join the call or reject it
- The official response from the White House and Beijing
- The timing of Anthropic's IPO relative to its pause rhetoric
- Whether the paper remains a statement or turns into a concrete regulatory proposal
- The reaction within the EU as it revises the rules of its AI Act
Also read: What Europe is writing about: The House approved aid to Ukraine despite Trump, Zelensky wrote to Putin, sanctions on Havana
