.jpg)
The TechLens
Artificial intelligence is advancing at a pace that is reshaping technology, business and scientific research. But alongside the rapid development of increasingly capable AI systems, a growing number of prominent technology leaders and researchers are calling for stronger safeguards, independent oversight and international cooperation.
In recent weeks, OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, Google DeepMind CEO Demis Hassabis and AI pioneer Yoshua Bengio have each raised concerns about the risks associated with increasingly capable AI systems, although their approaches and emphasis differ.
The debate reflects a fundamental question facing the technology industry: How can AI development continue at speed while ensuring that increasingly powerful systems remain safe, controllable and accountable to humans?
Sam Altman: AI Must Remain Under Human Control
OpenAI CEO Sam Altman has increasingly emphasized the importance of maintaining human control as AI systems become more capable.
Speaking at the United Nations Security Council in September 2026, Altman joined other AI leaders in discussing both the enormous potential and risks of frontier artificial intelligence. The discussions focused on the possibility that advanced AI could create security challenges that individual companies or countries may not be able to manage independently. (Reuters)
Altman has argued that AI could produce major advances in areas such as scientific discovery, productivity and human capability. At the same time, he has acknowledged that society must take seriously the possibility of losing control over highly capable systems.
For Altman, the challenge is not simply building more powerful models. It is developing the technical safeguards and governance structures necessary to ensure that those systems remain aligned with human objectives.
His recent position also highlights another concern: the concentration of AI power. If extremely capable AI systems are controlled by only a small number of companies, individuals or countries, the technology could create new forms of economic and geopolitical influence.
Dario Amodei: AI Development Needs to Slow Down
Anthropic CEO Dario Amodei has made one of the clearest recent calls for a change in the pace of frontier AI development.
In September, Amodei argued that AI companies should “slow the pace” of development so that safety measures can catch up with rapidly increasing capabilities. His proposal includes greater access for independent safety evaluators, coordination between frontier AI companies and stronger international cooperation. (Reuters)
Amodei has not called for abandoning AI development. Instead, his argument is that capability growth should be accompanied by adequate testing, evaluation and safeguards.
He has warned that AI systems are becoming increasingly capable of performing complex tasks autonomously. That creates new questions around cybersecurity, misuse, fraud and the ability of humans to understand and control system behaviour.
His proposal places independent evaluation at the centre of AI safety, suggesting that companies developing frontier systems should not be the only entities responsible for determining whether those systems are safe.
Demis Hassabis: A Global AI Watchdog
Google DeepMind CEO Demis Hassabis has taken a different but related approach.
In July 2026, Hassabis called for a new U.S.-led global AI watchdog or standards body capable of evaluating the world’s most advanced AI models. He proposed a framework under which frontier AI laboratories could voluntarily submit models for assessment before release, with the possibility of formal requirements developing later. (Axios)
The proposal reflects a growing recognition that frontier AI may require oversight mechanisms similar to those used in other high-impact industries.
Hassabis has long argued that AI could deliver enormous benefits, particularly in scientific research and medicine. His position therefore combines optimism about AI’s potential with the view that increasingly capable systems require systematic evaluation.
The central idea is straightforward: AI companies should not operate without common technical standards for assessing the risks of their most advanced models.
Yoshua Bengio: Three Global AI Risks
AI pioneer Yoshua Bengio took the discussion to the United Nations Security Council on September 23.
Bengio identified three major categories of global risk: catastrophic misuse, concentration of power and loss of control over AI systems. He argued that no individual country can address these challenges alone because advanced AI technologies and their potential consequences can cross national borders. (Yoshua Bengio)
Bengio called for independent scientific assessments, common definitions for AI incidents and mechanisms for reporting serious security events.
He also proposed that frontier AI should be subject to a stronger international framework, including licensing and accountability measures comparable in principle to those used for other critical technologies.
His message was broader than corporate AI safety. He argued that decisions affecting the future of advanced AI should involve governments, independent experts and societies rather than being determined primarily by a small number of companies and countries.
Mustafa Suleyman: The Control Problem
Microsoft AI CEO Mustafa Suleyman has also spoken about the difficulty of controlling increasingly autonomous AI systems.
The concern is particularly relevant as AI moves beyond conventional chatbots toward systems capable of using tools, interacting with software, carrying out multi-step tasks and operating with greater autonomy.
Microsoft CEO Satya Nadella has similarly emphasized that AI should remain under human control, reflecting a broader concern within the technology industry about ensuring that increasingly capable systems continue to serve human purposes. (AP News)
This focus on control is becoming increasingly important as AI agents move into enterprise environments. An AI system that can independently access applications, execute instructions or make decisions can create substantially different risks from a system that simply generates text in response to a prompt.
Not Everyone Wants to Slow AI
The AI industry is not united behind a slowdown.
Nvidia CEO Jensen Huang has pushed back against arguments for significantly slowing AI development, emphasizing the technology’s potential and describing safety as an engineering challenge. Other technology executives have also expressed concerns that excessive restrictions could limit innovation and economic opportunity. (The Wall Street Journal)
This disagreement is important because it demonstrates that the current AI debate is not simply a choice between “AI progress” and “AI safety.”
The larger question is how the two can develop together.
Technology companies are competing to build increasingly capable systems, while governments and researchers are attempting to understand what safeguards are necessary. Finding the appropriate balance between innovation, security and accountability remains an open policy and technical question.
Why AI Safety Has Become a Boardroom Issue
The discussion is no longer limited to AI researchers.
For businesses, increasingly autonomous AI systems could transform software development, customer service, cybersecurity, finance, healthcare, manufacturing and knowledge work. But greater autonomy also means greater responsibility.
Companies adopting advanced AI will need to consider issues such as data security, model reliability, access controls, human oversight, auditability and incident response.
For boards and technology leaders, AI governance is therefore becoming part of enterprise risk management rather than simply an IT issue.
The emergence of AI agents makes this particularly significant. An AI system that can independently perform tasks can potentially create value at a much larger scale, but an error or misuse can also propagate rapidly across connected systems.
From AI Race to AI Governance
The statements from Altman, Amodei, Hassabis, Bengio and Suleyman reveal a significant evolution in the AI conversation.
A few years ago, much of the discussion centred on what generative AI could accomplish. Today, the conversation increasingly includes questions about who controls AI, how advanced systems should be tested, who is responsible when systems cause harm, and what international rules should govern frontier models.
There is no single consensus among technology leaders on the right approach.
Some are calling for slower development. Others favour independent standards and testing. Some emphasize technical solutions, while others focus on international governance and accountability.
But there is growing recognition that increasingly capable AI systems cannot be treated simply as another software upgrade.
The Road Ahead
Artificial intelligence has the potential to accelerate scientific discovery, improve productivity and create new products and services. At the same time, its increasing autonomy introduces risks that may be difficult for individual organisations to manage alone.
The recent warnings from leading AI figures therefore represent an important shift in the industry’s conversation.
The central challenge is not whether society should benefit from AI. It is how those benefits can be achieved while maintaining meaningful human oversight, responsible development and appropriate safeguards.
As AI capabilities continue to advance, the next phase of the industry may be defined not only by who builds the most powerful models, but also by who can demonstrate that those models can be developed and deployed responsibly.
For businesses, governments and technology leaders, AI safety is increasingly becoming inseparable from AI strategy.
The race to build the next generation of AI is continuing. But alongside that race, a second effort is gaining momentum: building the systems, standards and international cooperation needed to ensure that increasingly powerful AI remains accountable to the people it is designed to serve.
