AI Safety Risks: What Anthropic CEO Dario Amodei Says About Advanced Artificial Intelligence

Anthropic CEO Dario Amodei has renewed his call for the artificial intelligence industry to slow the pace of developing increasingly powerful AI systems, warning that the risks will depend heavily on the choices made by companies and governments. In an interview with CNN's Anderson Cooper, Amodei said the possibility of AI causing serious harm should not be viewed simply as a fixed probability, but as an outcome shaped by decisions taken along different paths.

Amodei's remarks came after Jacob Coxon, an AI researcher who previously worked at Anthropic and OpenAI, resigned and raised concerns about the industry's approach to developing advanced systems. Addressing whether AI could eventually kill humans, Amodei said he agreed with Coxon's concerns more than he disagreed with them. He said Coxon's criticism was directed less at Anthropic specifically and more at what he described as an industry-wide tendency to move too quickly.

The Anthropic CEO argued that discussions about AI risk are often reduced to numerical estimates, which can make the issue appear similar to a gamble. Instead, he said the industry should examine the different paths that AI development could take and focus on making decisions that reduce the likelihood of dangerous outcomes. Amodei stressed that companies and governments have an ability to influence those outcomes through cooperation, safety measures and responsible development.

Also Read: Anthropic Researcher Resigns After Warning AI Could Pose Deadly Risks Before 2030

In a recent blog post, Amodei similarly warned that AI could bring serious risks alongside its potential benefits. He identified concerns including losing control of advanced AI systems, the misuse of AI for cyberattacks and bioterrorism, as well as major economic disruption. He also warned that commercial competition could intensify those risks if companies prioritize speed over adequate safeguards, arguing that the industry should create more time to strengthen safety measures while continuing to make progress.

Amodei's call for caution has received support from other prominent technology leaders. OpenAI CEO Sam Altman said his company agreed that the development of frontier AI needs to be paced and backed the idea of independent evaluators having employee-like access to assess advanced systems. Tesla and xAI CEO Elon Musk also supported Amodei's concerns, writing that Amodei was right. Meanwhile, UN human rights chief Volker Turk has urged countries to establish strong guarantees for the safety and security of AI before its risks become more difficult to manage.

The debate reflects growing concern over how quickly AI capabilities are advancing and whether safety systems are developing at the same pace. Amodei maintained that slowing development does not mean abandoning technological progress, but rather using the additional time to strengthen safeguards and encourage cooperation among AI companies and governments. The central question, he argued, is therefore not simply how likely AI-related harm may be, but whether those developing and regulating the technology make the choices needed to keep its risks under control.

Also Read: Anthropic Uncovers UAE-Linked Campaign Aimed At Shaping Narratives Around Sudan War