AI Is Moving Too Fast: Why Global Leaders Are Calling for a Safer Pace

 

Artificial intelligence is advancing at extraordinary speed.

In only a few years, AI has moved from generating simple text and images to writing software, conducting research, analysing complex information, assisting with cybersecurity and increasingly acting on behalf of users through autonomous agents.

But in September 2026, the conversation around AI changed noticeably.

Some of the people building the world's most powerful AI systems are now warning that the technology may be progressing faster than governments, companies and society can safely manage.

The question is no longer simply:

How powerful can AI become?

It is increasingly:

How quickly should we allow it to become that powerful?

Why are AI leaders suddenly talking about slowing down?

On 12 September 2026, Anthropic CEO Dario Amodei called on leading AI companies to slow the rate at which they increase the capabilities of their most advanced models.

His argument was not that artificial intelligence should stop developing altogether.

Instead, he proposed giving researchers, governments and safety teams more time to understand and control the risks created by increasingly capable systems.

His framework reportedly includes independent safety evaluation, cooperation between leading AI laboratories and eventually greater international coordination around advanced AI development.

That distinction is important.

This is less about stopping AI and more about pacing frontier AI — the most powerful systems being developed at the limits of current technology.

OpenAI and Elon Musk have supported the discussion

The debate became much more significant when other major figures in the technology industry supported Amodei's concerns.

OpenAI CEO Sam Altman and Elon Musk have both backed the broader idea that the development of the most advanced AI systems may need to proceed more cautiously.

OpenAI has also recently called for stronger AI regulation.

On 9 September, the company said it supports mandatory national AI safety requirements based on the capabilities of advanced models.

It also expressed support for independent safety assessments and safeguards addressing areas including biological threats and risks to young people.

This is a significant shift.

The debate about AI safety is no longer being driven only by governments, academics or technology critics.

Increasingly, it is also coming from the organisations building the technology.

What has changed?

The capabilities of frontier AI systems are becoming significantly more powerful.

OpenAI recently reported that its Astra model had reached what it calls a Critical cybersecurity capability threshold.

According to OpenAI's assessment, a model at this level may be able to discover previously unknown security vulnerabilities and develop methods to exploit sophisticated computer systems with less direct human supervision.

OpenAI said it delayed parts of Astra's development and release while additional protections were put in place.

This demonstrates the challenge facing the industry.

The same AI systems capable of helping cybersecurity professionals find vulnerabilities can potentially provide powerful capabilities to malicious actors.

This is what is often referred to as the dual-use problem.

A technology can be extremely beneficial while simultaneously creating new risks.

Cybersecurity is becoming one of the biggest concerns

Anthropic has also reported increasingly sophisticated attempts by unauthorised organisations to obtain access to powerful US AI models.

Its September threat-intelligence report describes groups using proxy services, stolen API credentials and false identities to bypass geographic restrictions and access frontier AI capabilities.

This highlights an important reality.

AI safety is not only about science-fiction scenarios involving machines becoming uncontrollable.

Some of the risks already exist today.

They include:

• cybercrime;                         
  
• fraud;
• manipulation;
• misinformation;
• data security;
• biological misuse;
• abuse of autonomous AI agents.

As AI becomes more capable, the potential impact of misuse also increases.

Does this mean AI development will actually slow down?

Not necessarily.

One of the biggest obstacles is international competition.

Countries increasingly view artificial intelligence as strategically important for economic growth, defence, science and national security.

This creates what could be described as an AI race.

If one country voluntarily slows its development, it may worry that another country will continue advancing.

That tension became clear this weekend.

US President Donald Trump rejected some of the growing concerns around AI development and emphasised the importance of maintaining American leadership in artificial intelligence.

China has also pushed back strongly against the proposed slowdown.

A Chinese state-backed newspaper criticised Amodei's proposal, arguing that restrictions could become a way of limiting China's technological development rather than purely addressing safety.

This reveals the central problem.

AI safety requires cooperation, while AI competition encourages countries and companies to move faster.

Those two forces are increasingly colliding.

The financial markets are paying attention

Investors are also beginning to consider what stricter AI safety requirements could mean for the technology industry.

On 14 September, shares of several AI-related companies declined after the warnings from major AI leaders.

Companies connected to semiconductors, AI infrastructure and large-scale computing were among those affected.

This does not mean that investment in AI is disappearing.

Far from it.

Global spending on AI infrastructure remains enormous.

But investors are increasingly asking another question:

Will future AI growth be limited by safety requirements, regulation or slower frontier development?

That uncertainty is beginning to influence the market.

A slowdown does not mean stopping innovation

This point is especially important.

Calls to slow frontier AI development should not be interpreted as a call to abandon artificial intelligence.

AI already provides valuable benefits across many industries.

It is helping researchers analyse scientific data.

It is assisting doctors and healthcare professionals.

It is improving cybersecurity.

It is helping businesses automate repetitive tasks.

It is supporting education, software development and productivity.

The challenge is to make sure that capability does not advance significantly faster than our ability to manage it responsibly.

A sensible approach could therefore involve:

stronger safety testing before powerful models are released;

independent evaluation of frontier systems;

better cybersecurity controls;

transparent reporting of serious incidents;

clear legal accountability;

greater international cooperation;

and continued investment in AI safety research.

The goal would not be to prevent innovation.

It would be to ensure that innovation remains manageable.

Why human oversight still matters

As AI systems become increasingly autonomous, human responsibility becomes even more important.

AI may be able to analyse information, make recommendations or perform tasks extremely quickly.

But people and organisations still need to determine:

Should this system be trusted?

Who is responsible if something goes wrong?

What information should an AI system be allowed to access?

Which decisions should always involve a human?

These questions cannot be solved by better algorithms alone.

They require governance, ethics, cybersecurity, regulation and leadership.

Businesses should not panic — but they should prepare

Most organisations do not need to stop using AI because of the current debate.

However, companies should avoid adopting artificial intelligence simply because it is fashionable.

Responsible adoption means understanding:

what the technology is being used for;

what data it can access;

what risks it introduces;

who remains accountable;

how employees are trained;

and how the organisation protects sensitive information.

For many businesses, AI governance may soon become just as important as AI adoption itself.

The next phase of AI may be about control

For several years, the technology industry has largely focused on making AI models faster, smarter and more capable.

The next phase may require something different.

Control.

How do we ensure increasingly capable systems remain secure?

How do we prevent misuse?

How do we protect individuals and businesses?

And how do countries cooperate when they are simultaneously competing for technological leadership?

There are no simple answers.

But the fact that some of the world's leading AI developers are now asking these questions publicly is significant.

So, does the world need to slow AI down?

There is currently no global agreement to stop or broadly slow artificial intelligence.

Different governments, companies and researchers strongly disagree about how serious the risks are and how aggressively they should respond.

But something important has changed.

The conversation is moving from:

“How quickly can we develop AI?”

toward:

“How quickly can we develop AI safely?”

That may ultimately be the more important question.

Artificial intelligence has enormous potential to improve productivity, healthcare, education, scientific research and countless other areas of life.

But powerful technology requires equally powerful responsibility.

The future of AI should therefore not be a choice between innovation or safety.

It should be about finding a way to achieve both.

Final thought

Perhaps AI does not need to stop.

Perhaps it simply needs enough time for human judgement, regulation, cybersecurity and responsibility to keep up.

About the Author

Isabel Swart

Training Coordinator | Skunkworks Academy | South Africa

Through Skunkworks Academy, Isabel Swart is committed to making high-quality technology education accessible, practical, and aligned with real industry needs, empowering professionals to build confidence, advance their careers, and embrace lifelong learning.

📧 Email:isabel@skunkworksacademy.com

Skunkworks is all about combining technology, skills, and practical solutions to help people and businesses grow in a fast-changing digital world. From IT solutions to training and online learning, the focus is on creating real opportunities through technology.