AI Is about to escape human control — and nobody has a plan

June 13, 2026 - 21:48

The artificial intelligence company Anthropic has issued a stark warning to the world: advanced AI systems are rapidly approaching a point where they will operate beyond human oversight, and no effective plan exists to stop it. In a new policy paper, the firm calls for an immediate international pause on the development of the most powerful AI models, arguing that current safety measures are dangerously inadequate.

Anthropic's researchers point to recent breakthroughs in "agentic" AI-systems that can set their own goals, take independent actions, and even rewrite their own code. These capabilities, the company argues, are evolving faster than the guardrails designed to contain them. Without a coordinated global moratorium, they say, humanity risks creating machines that cannot be reliably controlled or turned off.

The proposed pause would apply to any AI system that demonstrates the ability to autonomously improve itself or operate in the physical world without direct human approval. Anthropic acknowledges that such a halt would be difficult to enforce, but insists it is the only realistic path forward. The company warns that the alternative is a future where AI escapes human control, with consequences that no one can predict or manage.