Should We Take Tech CEOs Seriously?
As Artificial Intelligence advances at a remarkable pace, the industry’s leading figures have warned of the existential threats. But how much of this concern is supported by actual scientific evidence?
In recent years, Artificial intelligence (AI) has become increasingly capable, transforming a variety of fields from scientific research to media and design. However, the speed of its development has raised concerns amongst some of the most influential figures. In September 2026, Anthropic chief executive Dario Amodei called for a “slower pace of advanced AI development”. His message received support from his rival and OpenAI CEO Sam Altman. Elon Musk chimed in, writing on X: “Dario is right.” Their warnings raise an important question: should society trust the people building AI to assess its dangers?
Concerns about AI extend beyond hypothetical scenarios involving machines surpassing human intelligence. The International AI Safety Report 2026, led by computer scientist Yoshua Bengio, identified risks associated with malicious use, system failures and wider societal disruption.
Some of these issues have already been documented. AI-generated content can be used to facilitate non-consensual imaging. Increasingly capable systems raise concerns about potential misuse in cybersecurity or for military use. On top of that, AI can produce inaccurate information, creating risks when blindly trusted.
However, more extreme predictions remain less certain. Researchers are investigating whether future AI systems could act in ways that are difficult for humans to monitor. Although these possibilities do require attention, their likelihood and severity remain the subject of scientific debate. Therefore, distinguishing demonstrated harms from potential future threatens is essential to an evidence-based discussion and decision.
The report highlights a central difficulty: AI capabilities are developing rapidly, but reliable evidence about some of their consequences is taking longer to establish.
This does not prove that catastrophic outcomes are inevitable, but it demonstrates why uncertainty shouldn’t automatically be interpreted as safety. Waiting for conclusive evidence of a serious threat could leave us with too little time to respond. Equally, treating every worst-case theory or predication as a fact could lead to disproportionate restrictions based on uncertain assumptions.
A scientific approach should therefore prioritise continuous independent assessment of AI systems as it continues to change. The aim should be to establish what risks can be accurately measured, what remains uncertain and what safeguards are justified by available evidence.
Altman, Amodei and Musk have substantial experience in the technology industry. They have access to information about the advancement of AI that nobody has access to. Their influence permeates beyond Silicon Valley. Amodei and Altman were invited to this year’s G7 summit in France. At last month’s UN General Assembly, they participated in a round table on AI. And let’s not ignore Musk’s strong ties to the Trump administration, having spearheaded the Department of Government Efficiency (DOGE). Even so, we shouldn’t dismiss warnings just because they come from business leaders. However, expertise does not make their predictions infallible. Their commercial positions also deserve scrutiny. AI companies have a vested interest in attracting investment, maintain public attention, whilst keeping an eye on evolving policy. Warnings about dangerous AI could encourage stricter oversight, potentially affecting competitors differently. Meanwhile, promoting AI’s capabilities attracts customers and funding. These contradicting incentives do not prove dishonesty, but they do reinforce the need for independent investigation.
Crucially, the views of these executives should not be confused with a scientific analysis on potential future outcomes. Independent researchers, policymakers and the public must also evaluate claims according to objective evidence, rather than the status of the person creating them.
Ultimately, their warnings do deserve attention, but not blind acceptance. The documented harms of AI already justify stronger safeguards. Uncertain long-term threats require continued research rather than predictions. Governments, researchers and companies should work together towards independent evaluations and greater transparency to society. The challenge is to ensure that the development of increasingly powerful AI remains accountable to the society it is intended to serve.
Words by Evie Conforth