Some of the world's most prominent artificial intelligence executives are calling for a more cautious approach to the development of increasingly powerful AI systems, in an unusual alignment between leaders of rival companies.
Dario Amodei, chief executive of Anthropic, the company behind Claude, has called for the AI industry to deliberately slow the pace at which frontier models are becoming more capable.
Amodei argues that AI capabilities are advancing faster than researchers and policymakers can develop adequate safeguards. He says the objective should not be to stop AI development, but to give safety measures enough time to catch up with the technology.
His proposal has received support from OpenAI CEO Sam Altman and Elon Musk, whose companies are among those competing to develop increasingly advanced AI systems.
Altman has previously acknowledged that AI development may need to be paced as models become more capable. He has argued that society needs sufficient time to strengthen its systems and institutions before deploying technologies with significantly greater capabilities.
Musk, a longtime critic of insufficient AI safeguards and the founder of xAI, has also endorsed Amodei's call for caution. The agreement is notable because the three executives represent competing companies operating at the forefront of the AI industry.
Concerns over increasingly autonomous AI
The debate is being driven by concerns about what could happen as AI systems become capable of performing increasingly complex tasks with limited human supervision.
Amodei has warned that future AI systems could potentially coordinate large numbers of autonomous agents and gain powerful capabilities in areas such as cybersecurity. The Associated Press reported that he believes AI could reach a level within six to 12 months at which coordinated AI agents could potentially control significant parts of the internet.
Such predictions remain forecasts rather than established facts. The timing and extent of so-called artificial general intelligence or superintelligence remain highly contested among researchers.
Nevertheless, recent incidents involving AI systems have intensified concerns within the industry.
Reports have highlighted cases in which AI agents demonstrated unexpected abilities involving cybersecurity and interactions with external systems. These incidents have prompted researchers and executives to question whether existing testing and containment measures are sufficient for increasingly capable models.
Amodei's proposed safeguards
Rather than advocating a blanket ban on AI development, Amodei has proposed a three-part approach.
The first element involves independent safety evaluators being given meaningful access to advanced AI systems so they can assess whether companies are meeting agreed safety standards.
The second calls for greater coordination between AI companies, including the development of common safety standards.
The third involves international cooperation, with governments working together to establish safeguards around increasingly powerful AI systems.
Amodei has also stressed that slowing the development of frontier AI should not mean abandoning technological progress or allowing other countries to gain an unchecked advantage.
A difficult balance
The proposal highlights a growing tension within the AI industry.
Companies have strong commercial incentives to build more powerful models, while governments view advanced AI as strategically important. At the same time, researchers and policymakers are increasingly concerned that technological capabilities could advance faster than safety mechanisms.
Altman has previously warned that attempts to slow AI development must also avoid creating systems in which only a small number of companies or governments control the technology. He has argued that legitimate safety concerns should not become a justification for excessive concentration of power.
The disagreement is therefore not simply between people who support AI and those who oppose it. Much of the current debate concerns how quickly AI should advance, who should oversee it and what safeguards should be required before increasingly capable systems are deployed.
For now, there is no agreement among governments or technology companies on a specific global limit for AI development. But the public alignment between Amodei, Altman and Musk represents a significant development in the industry's debate over how to manage the next generation of AI.
The central question is increasingly becoming not whether AI development should continue, but whether the technology can advance quickly enough to deliver its promised benefits while giving researchers, governments and society enough time to manage the risks.
