Sam Altman Warns Against AI Takeover, Says Humans Must Remain in Control of Artificial Intelligence

166
20 Nov 2026
min read

News Synopsis

OpenAI CEO Sam Altman has warned against allowing artificial intelligence to control humanity's future, stressing the need for stronger safety measures and continued human decision-making.

Sam Altman Raises Concerns About the Future of AI Control

The rapid development of artificial intelligence is creating new opportunities across industries, but it is also raising serious questions about how much control humans should give these increasingly powerful systems. OpenAI CEO Sam Altman and company president Greg Brockman have warned that allowing AI to gradually take control of important decisions could create significant risks for society.

Speaking at The New York Times' DealBook Summit, the two OpenAI leaders discussed the challenges associated with developing more advanced AI models. They emphasised that humans must remain responsible for shaping the future rather than allowing artificial intelligence systems to make decisions independently.

Their comments come amid growing discussions in Silicon Valley about the possibility that highly capable AI systems could eventually manage complex operations across businesses, institutions and other areas of society.

Altman and Brockman believe the technology industry needs to reconsider how it develops and deploys AI. According to their concerns, maintaining human oversight will require stronger security practices, improved monitoring and a more responsible approach to technological competition.

Sam Altman Rejects the Idea of Handing Over Control to AI

One of Altman's main concerns is the possibility of humans gradually surrendering decision-making authority to artificial intelligence. He and Brockman disagree with the idea that increasingly advanced AI systems should eventually take charge of important processes.

A major issue connected to this concern is known as recursive self-improvement (RSI). This refers to a situation in which an AI system helps develop or improve newer versions of itself, potentially creating systems that become more capable over time.

If this process continues without effective human supervision, controlling the direction of AI development could become increasingly difficult.

Altman believes companies must prepare for this possibility rather than assuming that existing development practices will remain sufficient. He also pointed out that competitive pressure between technology companies could make the situation more complicated.

Organisations may feel encouraged to develop increasingly powerful models quickly to remain ahead of competitors. However, prioritising speed without adequately considering safety could create problems.

During the summit, Altman warned that the next phase of AI development cannot follow the same approach as the previous few years.

He said, "If the world and the companies developing AI treated the next four years like they've treated the last four years, I think that would be very bad."

His statement reflects the need for a more cautious strategy as artificial intelligence becomes increasingly capable.

AI Safety Measures Must Advance Alongside Technology

Another major issue discussed at the summit was whether existing safety and monitoring systems can keep up with the growing capabilities of advanced AI models.

As artificial intelligence becomes more sophisticated, understanding how models reach particular conclusions is becoming increasingly important. Developers need reliable methods to examine their behaviour, identify potential problems and ensure that systems operate within appropriate limits.

Greg Brockman highlighted concerns about the effectiveness of existing chain-of-thought monitoring techniques.

Why Chain-of-Thought Monitoring May Become Less Effective

Chain-of-thought monitoring involves examining the reasoning steps that certain AI models produce while working through a problem. Researchers use these outputs to gain insights into how models approach tasks and arrive at answers.

However, Brockman suggested that this approach may become less reliable as advanced models perform more internal computation without explicitly displaying their reasoning process.

If AI systems increasingly carry out complex operations without showing their complete thought process, researchers could find it harder to understand what is happening inside them.

This creates a challenge for AI developers because monitoring systems must evolve alongside the technology. Companies may need additional methods to evaluate model behaviour, identify risks and maintain appropriate oversight.

Sam Altman Calls for Shared AI Safety Standards

Altman also emphasised the importance of cooperation across the artificial intelligence industry. He believes companies should establish stronger and more consistent safety practices instead of developing advanced systems without sufficient understanding of the risks involved.

AI development is becoming increasingly competitive, with major technology companies investing heavily in more powerful models. However, safety concerns cannot be addressed effectively if organisations focus only on their individual development targets.

Shared safety standards could help companies evaluate risks, improve security practices and establish clearer expectations for responsible AI development.

The discussion also highlights the importance of transparency and communication between organisations working on advanced artificial intelligence.

According to the concerns raised by Altman, the industry needs to understand how different companies are developing their systems and whether their safety measures are keeping pace with technological progress.

OpenAI Redirects Engineering Teams Towards Security

The discussion about AI safety also comes after a security incident involving an OpenAI model and Hugging Face.

Brockman revealed that OpenAI had temporarily paused 25% of its production engineering projects during a security-focused sprint. Engineers were redirected towards strengthening the company's systems with the help of AI models.

This decision demonstrates how security concerns can influence the priorities of an AI organisation, even when it is working on several development projects.

The incident also highlights the growing importance of protecting AI infrastructure as models become more powerful and widely used.

Companies developing advanced AI systems must consider not only how their products perform but also how securely they are built, deployed and maintained.

Altman Says Humans Must Retain Their Independence

Beyond technical risks, Altman raised a broader concern about the impact of artificial intelligence on human freedom and decision-making.

He argued that people should not become dependent on AI systems to the extent that they lose the ability to make important choices or shape their own lives.

During the summit, Altman used an unusual comparison to explain his position, saying, "I don't want to be a Labrador. I want my kids to be Labradors."

His remark referred to a possible future in which humans could become less independent as increasingly capable AI systems take on more responsibilities.

Altman stressed that people should continue to have the freedom to make decisions, create new things and determine their own future.

The underlying message is that artificial intelligence should support human capabilities rather than replace human agency. Even if AI becomes significantly more advanced, people should remain involved in decisions that affect society.

Conclusion: Human Oversight Will Be Crucial as AI Advances

Sam Altman and Greg Brockman's comments highlight the growing importance of AI safety, security and human oversight. As advanced models become more capable, existing monitoring methods and development practices may need significant improvements.

Concerns about recursive self-improvement, competitive pressure and reduced human agency show why responsible AI development requires careful planning. The challenge for technology companies will be to continue innovating without allowing safety and human control to become secondary priorities.

Podcast

TWN Exclusive