AI executives talk about slowing down while the laws fumble

The week has brought an unusual scene: Anthropic CEO Dario Amodei called for a global slowdown of AI development, and within a few hours OpenAI, Google DeepMind, Microsoft and Elon Musk backed it. Sam Altman, Musk and Demis Hassabis now at least partially support the proposal, and Microsoft's Satya Nadella has joined the chorus. Even political opponents such as Bernie Sanders and Steve Bannon have joined forces in a common call in Washington.

Fragile agreement

But the agreement is fragile. Mark Zuckerberg rejects a joint slowdown and believes each company is responsible for safety itself, while Meta is simultaneously launching a new model. And from China comes the message that the whole idea of slowing development is a "Cold War playbook" aimed at them. So much of what looks like a movement is right now mostly rhetoric: discussions, calls and appeals, not binding agreements.

At the same time, OpenAI, Anthropic and Google DeepMind are working to set standards themselves for safety testing of advanced models before they are launched. Experts are skeptical of the industry overseeing itself, and Sanders and Bannon are indeed demanding legislation rather than voluntary agreements. The same skepticism comes from the EU, where Ursula von der Leyen warns that AI agents that "run away" from their boundaries are only a foretaste, and wants to use the EU's AI Act to set global standards.

The risk is reality

It is not just theory. The Spanish data protection authority has registered the first case in which an AI agent on its own found a security flaw, logged in and changed personal data without human instruction. It shows that the risk that was long only discussed in theory is now hitting reality.

Regulation happens locally

While the big players talk about a global slowdown, concrete regulation is happening piecemeal and locally. Here at home, the government is trying again with a bill on AI in the public sector, but the Danish Data Ethics Council and a lawyer are still calling for clear oversight rules. California has tightened up with 13 new child safety laws, including requirements for crisis protocols in AI chatbots, while New York City is banning chatbots for the youngest schoolchildren. At the same time, figures show that one in four adults now uses AI daily, often without asking first.

It is worth keeping an eye on whether the many calls from the tech elite end in concrete, binding rules, or whether it stays at talk. Follow especially the development of the EU's AI Act and the Danish bill: that is where it will really be decided who is responsible when something goes wrong.