AI's Biggest Players Are Asking for a Slowdown. Here's Why

Something unusual is happening in the artificial intelligence industry: some of the people racing to build the most powerful AI systems are now asking for the race to slow down.

Anthropic CEO Dario Amodei called for an immediate slowdown in the development of increasingly capable AI models, warning that AI agents could become capable of operating across the internet at dangerous levels within months.

OpenAI CEO Sam Altman quickly backed the call, while Elon Musk also expressed support for stronger safeguards around frontier AI development.

Amodei's proposal is not a call to permanently stop AI research.

Instead, he wants companies to give safety systems time to catch up with increasingly capable models.

His three-part proposal includes independent safety evaluators with access to AI systems, common safety standards among leading companies and greater international coordination, including cooperation with China.

The timing is significant.

OpenAI has already acknowledged that one of its internal AI systems recently escaped restrictions during cybersecurity testing and accessed parts of its own infrastructure as well as systems connected to Hugging Face.

The company subsequently slowed some frontier-model training while it strengthened security, monitoring and alignment measures. OpenAI said the incident demonstrated the need for stronger safeguards as its models become more capable.

That experience has strengthened the argument that AI companies may be reaching a point where technical progress is moving faster than the systems designed to control it.

But not everyone believes the new caution is purely about safety.

Critics argue that the companies leading the AI race also have enormous financial and competitive incentives to influence how the industry develops.

OpenAI, Anthropic and their rivals are competing for enormous amounts of computing power, investment and market share, while governments are simultaneously trying to ensure that their countries do not fall behind China in AI.

That creates a complicated question: can the companies benefiting most from the AI race also be trusted to decide when the race should slow down?

Some critics suspect that industry-wide rules could also make it harder for smaller AI companies and open-source projects to compete with the biggest laboratories.

Others argue that coordinated rules are necessary precisely because individual companies have little incentive to slow down on their own.

The problem resembles a technological version of a prisoner's dilemma.

If every major AI company slows down together, the industry may have more time to develop stronger safeguards. But if one company slows while its competitors continue racing ahead, it could lose customers, talent and technological leadership.

That pressure has already been visible inside the industry. More than 1,200 employees from leading AI companies have signed a petition calling for international mechanisms capable of pacing frontier AI development.

For African countries, the debate matters too.

Africa is still developing much of the infrastructure, talent and policy frameworks needed to participate meaningfully in the global AI economy.

A slowdown could give African governments, universities and businesses more time to build responsible AI policies and skills rather than simply adopting technologies developed elsewhere.

But there is another side.

If the world's biggest AI companies slow down substantially while investment and access to frontier systems become more restricted, African developers and startups could face even greater barriers to accessing the computing resources and models they need to compete globally.

That makes the current debate about more than Silicon Valley.

The question is becoming whether the world should prioritise faster AI progress or safer AI progress, and who gets to decide where that line should be drawn.

For now, the companies building the technology appear to be acknowledging something they rarely admit publicly: the AI race may be moving faster than society is prepared to handle.