Fears about runaway artificial intelligence are not new.

But they have gained fresh attention as warnings about rogue AI systems grow louder.

Several leading AI executives recently agreed on the need for a coordinated slowdown.

This includes international cooperation to protect humanity from worst-case situations.

Other tech leaders reject the idea that their companies cannot ensure safety on their own.

Anthropic CEO Dario Amodei has offered the most detailed slowdown plan.

He proposed that frontier labs give independent outside evaluators "ongoing, employee-like access."

This includes offices, badges and laptops.

Anthropic and OpenAI have both committed to this.

Amodei also called for government regulation and coordination among frontier AI companies.

He said democratic governments and authoritarian governments should also take part.

He acknowledged the difficulty, citing China in particular.

But he said "we owe it to humanity to try."

OpenAI CEO Sam Altman supports pacing development and more industry cooperation.

OpenAI has pushed for mandatory national safety requirements and backed state legislation.

Its chief global affairs officer, Chris Lehane, said the company would advocate for international approaches.

These would measure capabilities and manage risk, even if that slows model capabilities.

Altman said pacing does not mean stopping.

He said "no amount of American competitive pressure should justify recklessness."

Meta CEO Mark Zuckerberg pushed back.

He said each company should ensure its own technology is safe.

Nvidia CEO Jensen Huang also criticized slowdown calls.

He argued market forces are enough and no new laws are needed.

Elon Musk backed Amodei, writing "Dario is right."

He suggested competitors test each other's models.

Google DeepMind's Demis Hassabis agreed with Amodei's direction.

He proposed an industry-funded frontier AI standards body.

Microsoft's Satya Nadella said AI must help humanity and stay under human control.

But he warned the effort cannot be controlled by a handful of entities.

Microsoft AI CEO Mustafa Suleyman said he worries that anthropomorphizing models amplifies safety risks.