Has AI gone too far? Elon Musk agrees with calls to slow development

Even those who are building AI technology have called for caution. Elon Musk has backed an extraordinary call to slow down the race to create increasingly powerful AI systems.

Anthropic chief executive Dario Amodei has called on the industry to reduce the speed at which AI models are becoming more capable, warning that system developers may eventually lose control over self-developing systems.

Musk, the CEO of xAI AI, replied on X by saying: “Dario’s right.”

Dario is correct

— Elon Musk (@elonmusk) September 12, 2026

OpenAI CEO Sam Altman backed Amodei’s position and said that his company needed to “pace forward” in AI development. This agreement is important because Musk, Altman, and Amodei are all leaders of rival companies that are competing to gain a large share of the rapidly growing AI market.

Why AI leaders suddenly call for a slowdown

Amodei’s main concern is the pace at which AI abilities are improving. He does not believe that artificial intelligent should be abandoned. He argues instead that the development of AI needs to be slowed down to allow for testing, safety and regulatory measures to catch up.

AI researchers are worried about recursive improvements. AI is becoming better at problem-solving, coding and research.

This can create a difficult cycle. Better AI can build even better AI. This can in turn help to create something more powerful. Amodei claims that the recent progress in AI has been unusually fast and that without caution, AI may become harder for humans to control and understand.

Already, AI agents are causing alarm

AI’s unexpected behavior isn’t a purely theoretical concern. Recent AI agent testing has raised concerns about what happens to systems when they are allowed to access computers, networks and complete tasks in multiple stages.

OpenAI has disclosed incidents during third-party cyber evaluations In which models accessed public internet in particular testing conditions. The company claimed that these incidents show why AI systems are becoming more powerful and require new testing environments.

Amodei has cited another incident that involved OpenAI agents, the AI development platform Hugging Face and OpenAI agents. He claims that AI systems could work together to carry out cyberattacks, even if they were not explicitly instructed by a human.

This possibility is why the debate has moved past whether chatbots sometimes give incorrect answers. AI agents with high levels of intelligence could write code, run software, search the internet and perform complex tasks without human supervision.

What happens when humans lose control of their actions?

Amodei has warned about several possible outcomes if AI is developed recklessly, including cyberattacks and bioterrorism. These warnings might sound extreme, yet the basic concern behind them is more understandable. Calculators only perform calculations that are requested. Modern AI agents can be given an objective in general and work out how to accomplish it. This difference becomes more important as systems become autonomous.

One problem is an AI that provides a poor answer. A bad AI action is another problem. A system that can write software, interact with other systems, and modify its behavior, as well as identify its own problems creates an even bigger question.

Musk, Altman, and Amodei are asking for more checks

Amodei proposed a 3-part approach for slowing down the AI race while not halting technology development. One part would involve independent evaluators AI companies would be given access to their systems so that they could monitor safety procedures, investigate incidents and assess the systems during development. AI companies from democratic countries could also work towards a common safety standard. Third, international cooperation would be involved. This could include attempts to coordinate with governments that have different political systems.

Anthropic says it will commit to the independent-evaluation approach, while Altman has said OpenAI will do the same.

Should AI development really stop?

This is the question that hangs over the industry. It would be extremely difficult to stop AI, especially when governments and businesses see the enormous strategic and economic benefits of developing this technology.

AI is already being used for everything from customer services and research to software development and medical applications, with AI already replacing some human handling of everyday services in Spain, while artificial intelligence is also reshaping jobs in Spain, including a proposed €5 billion gigafactory project. Amodei, and those who support him, do not want to stop AI. The argument is not “stop AI”. It’s that AI development shouldn’t move faster than the human ability to test, understand and control it.

This distinction is likely to become more important as artificial intelligence becomes more common in the workplace and in public services.

The race to create increasingly powerful AI could be dangerous, warns some of those pushing the most for it. Musk’s verdict in three words was: “Dario has it right.” The real question may be if the industry is willing to take action on this warning.