SynopsisAnthropic chief Dario Amodei urges slowing advanced AI development for safety. He believes current progress outpaces our understanding and control capabilities. Recent incidents highlight risks of autonomous AI systems acting unexpectedly. Amodei proposes a framework for safety standards and international coordination. He advocates for careful development to realize AI's transformative benefits.AgenciesDario AmodeiAnthropic chief executive Dario Amodei has called for slowing the pace of frontier AI development, marking one of the strongest shifts yet from a leading AI lab chief who has previously argued for balancing rapid innovation with safety.In an essay published on Saturday, Amodei said recent advances have convinced him that artificial intelligence companies need to deliberately "pace the frontier" so safety research can keep up with increasingly capable models."We must slow the pace at which we improve the capabilities of AI models. Progress will still seem fast, and we must make wise use of the time we gain," he wrote.The essay comes days after two AI safety researchers at Anthropic -- Jacob Coxon and Evan Hubinger -- publicly argued that the industry's safety assumptions are being overtaken by the speed of model development. Both warned that recent AI systems are displaying increasingly autonomous behaviour and that companies may be approaching a point where capabilities outpace their understanding of how the models work.Amodei said AI is now advancing "drastically faster" because models are increasingly helping build the next generation of models, a phenomenon known as recursive self-improvement."Left unchecked, it could outrun our ability to understand and control these systems, and so must be pursued very carefully, if at all," he wrote.He also cited the recent OpenAI-Hugging Face incident, where autonomous AI agents reportedly carried out unintended cybersecurity attacks, attempted to manipulate evaluation systems and coordinated as a collective despite not being instructed to do so."A swarm that possessed greater capabilities but a similar level of misalignment could have caused catastrophic damage," Amodei wrote, warning that similar systems could become capable of taking over "the entire internet with a persistent botnet" within 6-12 months if safeguards do not improve.xAI chief Elon Musk concurred with Amodei's views. In a post on X citing Amodei's essay, Musk wrote, "Dario is right".To address the risks, Amodei proposed a three-stage framework centred on slowing capability gains rather than pausing AI altogether. Anthropic will begin by allowing embedded third-party evaluators to work inside the company with employee-like access to audit safety practices and publicly report findings.He also called for democratic governments to help frontier AI companies establish common safety standards and eventually pursue international coordination with countries including China.Despite advocating a slower pace, Amodei argued against halting AI development."Pacing does not mean halting model training or technical progress, but ensuring companies take adequate time to align and safeguard their models."He maintained that AI's transformative promise remains intact, reiterating his belief that it could "cure most major diseases in the next 5-10 years" and create unprecedented economic abundance."The benefits will only be achieved if we build the technology in the right way," he wrote. ...moreElevate your knowledge and leadership skills at a cost cheaper than your daily tea.Subscribe Now