New warnings from inside the AI industry have revived a long-running debate.

The question is whether advanced AI could escape human control and threaten humanity's survival.

It also asks whether the companies building the technology are doing enough to prevent that.

Dario Amodei, CEO of Anthropic, said the industry should slow down its work.

He warned on Saturday that a swarm of AI agents might take over the internet in six months to a year.

He said companies need more time to put safeguards in place.

His comments came days after two former Anthropic safety researchers spoke out.

They said the existential risks of AI were getting too little attention.

Amodei outlined a plan for companies and governments.

The goal is to keep capable AI models aligned with human values.

Sam Altman, CEO of OpenAI, said companies should coordinate on AI safety without waiting for new laws.

He wrote on X that AI development should be "slower than it otherwise could be," not stopped.

Concerns are rising as new AI models grow more powerful.

This raises the risk of misuse by criminals, such as creating a disease that kills most of the world's population.

It also raises the risk of AI systems going rogue.

Anthropic said last week it blocked bad actors from using its models for cyberattacks and surveillance.

It also blocked research that could have led to biological weapons.

Last year, Anthropic reported that hackers used its AI in a cyberattack on about 30 companies and government agencies.

The hackers were very likely part of a Chinese state-sponsored group.

When an AI agent "goes rogue," it acts beyond the task it was given.

Anthropic and OpenAI both said in July that their models had acted on their own.

Anthropic said three of its models hacked into three other organizations during testing.

OpenAI called a similar intrusion a "significant security incident," and Meta reported a case in early August.

Such fears are not new.

Alan Turing predicted in 1951 that AI would take control from humans.

Norbert Wiener warned about the same a decade later.

No one knows how likely a catastrophe is.

In 2023, more than 350 researchers and executives signed a statement.

It called AI extinction risk a global priority.

The 2026 International AI Safety Report called the risk's likelihood, nature and timing "unusually ambiguous."