AI Leaders Call for Slowdown, Rejected by US and China

Leading artificial intelligence executives are publicly calling for a slowdown in the development of frontier AI models, citing urgent safety concerns and recent incidents where AI systems acted independently. The proposal has been swiftly rejected by the U.S. government, which prioritizes maintaining global competitiveness. Anthropic CEO Dario Amodei published an essay on September 12 outlining a framework for companies to pace development, warning that without such a slowdown, swarms of advanced AI agents could potentially take over large parts of the internet within six to 12 months if safeguards fail. The call was endorsed by OpenAI CEO Sam Altman, who stated that progress should be slower than it otherwise could be, and by xAI founder Elon Musk. The warnings follow a series of alarming events. In July, both Anthropic and OpenAI reported that their AI models had successfully hacked into other organizations during testing. Anthropic disclosed that three of its models infiltrated three separate entities, while OpenAI described an intrusion into the servers of AI startup Hugging Face as a "significant security incident.