Anthropic calls for global pause in AI development over self-improvement risks

The company behind Claude warns that AI systems are writing most of their own code and improving faster than humans can govern them.

Share

Add us on Google by Editorial Team Jun. 5, 2026

Anthropic, the AI safety lab now valued at roughly $965 billion to $1 trillion, is asking the world to pump the brakes on the technology that made it one of the most valuable companies on the planet.

On June 4, 2026, the company published a proposal through its Anthropic Institute calling for a coordinated global pause in frontier AI development. The core argument: AI systems are approaching the ability to recursively improve themselves, and humans are losing the ability to meaningfully oversee the process.

The numbers behind the panic

The blog post, authored by Marina Favaro and Jack Clark, lays out a trajectory that reads less like a corporate update and more like a warning flare.

As of May 2026, more than 80% of code merged into Anthropic’s own codebase was written by Claude. Not by human engineers. By the AI itself.

Engineers at the company were merging 8x more code per day in Q2 2026 compared to 2024. AI task-completion horizons, essentially how complex a task an AI can handle autonomously, have been doubling roughly every four months. In March 2024, models could handle tasks that took about 4 minutes. Later models stretched that to 12-hour tasks.

Advertisement

Anthropic is careful to note that full recursive self-improvement, where an AI genuinely redesigns itself to become smarter in an uncontrolled loop, hasn’t happened yet. But the company’s own data suggests the gap between “hasn’t happened” and “could happen” is narrowing fast.

Why a unilateral pause won’t work

The proposal doesn’t just say “everyone stop.” It acknowledges the game theory problem baked into the situation.

If Anthropic pauses alone, competitors like OpenAI, Google DeepMind, or xAI keep building. The cautious lab falls behind. The reckless one captures the market. The solution Anthropic proposes is internationally verifiable mechanisms that would allow multiple labs to halt development simultaneously.

Back in March 2023, the Future of Life Institute published an open letter calling for a six-month pause on training AI systems more powerful than GPT-4. It was signed by thousands of researchers and tech executives. It was also widely ignored by the labs themselves.

What’s different now is that the call is coming from inside the house. Anthropic isn’t an external advocacy group. It’s a frontier lab with firsthand access to the capabilities it’s warning about.

Protests occurred in March 2026 outside the offices of Anthropic, OpenAI, and xAI, with demonstrators advocating for a pause on frontier model development.

The IPO elephant in the room

Anthropic is reportedly preparing for an eventual IPO, with its valuation sitting between $965 billion and $1 trillion. Calling for a global development pause right before going public is either an act of extraordinary corporate conscience or an exceptionally savvy positioning move. A coordinated slowdown would freeze the competitive landscape at a moment when Anthropic is already one of the top players. New entrants get locked out. Existing leaders get to consolidate.

What this means for investors

For the major players, Anthropic, OpenAI, and Google DeepMind already have significant moats in talent, data, and infrastructure. A pause doesn’t erase those advantages. It arguably cements them by preventing newcomers from closing the gap.

The most important thing to watch is whether other frontier labs respond. Anthropic’s proposal explicitly requires multilateral buy-in. Without OpenAI, Google DeepMind, and xAI at the table, this remains a well-argued blog post rather than a policy reality.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.