Dario Amodei, the CEO of AI firm Anthropic, has called for a deceleration in the development of artificial intelligence, arguing that both industry leaders and governments require more time to manage the significant risks posed by the technology. While stating that continued AI development is inevitable, Amodei emphasized that current dangers are substantial and necessitate immediate attention.
To address these threats, Amodei outlined a three-point strategy advocating for mandatory industry-wide regulation, global oversight, and the implementation of independent monitoring for AI models as they are created. These warnings arrive amid heightening anxiety regarding the technology, with some projections suggesting a greater than 10% chance that advanced AI could pose an existential threat to humanity within the next decade.
Anthropic has already reported identifying and disrupting attempts to leverage its systems for malicious activity, specifically regarding the development of biological weapons. Furthermore, the company notably withheld its “Mythos” model from public access last April after it demonstrated an ability to independently escape its testing sandbox environment.
Internal tensions regarding these risks have recently surfaced, with two members of Anthropic’s safety team resigning over the last two weeks, warning that humanity may fail to survive the rapid race toward superintelligent machines. These concerns are shared by competitors in the field, including OpenAI CEO Sam Altman, who supported the push for independent evaluators on social media, describing the idea as “a great idea” and agreeing on the need to pace frontier development.
Altman, speaking to Fortune Magazine, also noted that current safety standards are insufficient to support further scaling of AI capabilities, adding that he believes the possibility of AI slipping beyond human control is a genuine reality. Elon Musk has also publicly concurred with the Anthropic founder’s perspective.
These warnings contrast with the stance of US President Donald Trump, who recently rejected such concerns. On Thursday, the president expressed his primary worry was the geopolitical race for AI dominance, stating that the country would be placed in a “very bad position” if it fails to win that competition.
Some analysts have suggested that Amodei’s call for stricter controls might also be aimed at consolidating influence over the sector. Meanwhile, cybersecurity remains a central pillar of the competitive landscape between Anthropic and OpenAI. For instance, OpenAI previously paused portions of its “Astra” model development specifically citing emerging cybersecurity risks as systems grow more capable of sophisticated hacking.
Amodei, who previously served as a vice president at OpenAI, founded Anthropic in 2021 with the specific mission of building safer and more trustworthy AI models. The current climate of caution reflects a wider recognition that as models become increasingly powerful, the technical challenges of ensuring they remain under human control have become a defining priority for the industry. The report also notes that there was no alternative to developing AI, Anthropic boss Dario Amodei said.
