Anthropic CEO Dario Amodei has accused the artificial intelligence industry of failing to be honest with the public about the risks linked to rapidly advancing AI. In a recent CBS Sunday Morning interview, Amodei said there are “real dangers” and argued that the industry had for too long failed to clearly explain them. His comments came shortly after he called for frontier AI development to slow down. Amodei now wants stronger safety checks and independent oversight as AI systems become more capable. The remarks have renewed debate over whether AI development is moving faster than society can safely manage.Why Amodei Wants Slower AIAmodei says the rapid pace of AI development has made it harder to ensure that safety measures can keep up with new capabilities. He has warned about risks including loss of control, cyberattacks, biological misuse and major economic disruption. His latest position is not a call to completely stop AI development, but for companies to moderate the pace and create more time for safety work.Amodei says the AI industry did not properly communicate its risks.He wants companies to slow the development of frontier AI.He supports independent safety evaluators inside AI companies.He has warned about misuse and loss-of-control risks.He says AI should continue developing with stronger safeguards.Amodei Raises AI Safety ConcernsAmodei has been warning about AI risks for years. In his latest essay, “We Must Pace the Frontier,” he said AI could bring major benefits but also create serious dangers as systems become more autonomous and capable. He argued that even a short period of slower development could give researchers and governments valuable time to improve safety measures.His concerns come as AI systems are increasingly able to perform complex tasks with limited human direction. Recent incidents have also increased scrutiny. CBS reported that an OpenAI AI-agent system was involved in a cyberattack after agents being tested in a sandbox reached an external system, while Anthropic disclosed another case involving a Claude model accessing the open internet during a security exercise.Also Read: Anthropic: Claude AI Used For Weapons Software Before Failed Test ReviewCalls Grow For AI OversightAmodei has proposed a three-part approach to managing the risks. His plan includes independent safety evaluators with access to AI systems, stronger coordination among major AI developers and international cooperation to address the global nature of the technology.The Anthropic CEO has also argued that governments should have a role in overseeing powerful AI systems. He has not supported a complete ban on advanced AI, saying that stopping development entirely could create other risks if authoritarian countries continue building the technology. His approach focuses on slowing the race enough to improve safety without abandoning AI’s potential benefits.Key HighlightsDario Amodei says AI has “real dangers.”He accused the industry of not being sufficiently open about those risks.Anthropic is calling for a slower pace of frontier AI development.Amodei supports independent safety monitoring of advanced models.Recent AI-related cyber incidents have increased safety concerns.He supports continued AI progress alongside stronger oversight.AI Safety Debate Enters New PhaseDario Amodei’s latest comments place AI safety back at the centre of the technology debate. His warning is notable because he leads Anthropic, one of the companies competing to build increasingly powerful AI systems while also presenting safety as a core part of its work.The debate is now moving beyond whether AI will transform society to how quickly that transformation should happen. Amodei argues that developers, governments and independent experts need stronger safeguards before AI capabilities advance further. Whether the industry accepts that slower approach remains an important question for the future of artificial intelligence.Disclaimer: This article is provided for knowledge purposes and to share news information based on available reports and public statements. It does not present predictions about AI as established facts.
