Anthropic asks big AI companies to work together. It wants a pause in advanced AI work if risks rise.

In a blog post on Thursday, the company speaks. It says AI grows very fast now. It says the world should slow down. It says the world should also stop for a time.

Anthropic says its research group will work with others. They will look for ways to make a slowdown real. The company says a pause can give society time. Society can build better safety work and other guards.

It also says a shared system can help. It can show that rivals really stop. It can also stop a bad actor. That actor could race ahead in secret.

OpenAI takes a different view in a report. The report comes out on Wednesday. It says democratic governments should set the rules. It says private companies should not do it alone.

OpenAI says one lab should not decide. One company should not decide. A special interest group should not decide. It says the pace of new ideas needs wider control.

Anthropic says AI systems get faster at coding. It says they could one day design new ones. This is recursive self-improvement.

It says this could bring big good things. But it also says humans could lose control. The warning comes as researchers at the University of Toronto show more. They show how AI tools can help make a self-spreading “worm.” It changes its hacking plan across devices.