**[Anthropic Grants External Auditors Near-Employee-Level Access Following AI Extinction Debate]** Beating AI News Flash: Former Anthropic researcher Jacob Coxon's resignation post has garnered over 160 million views in just three days, propelling the topic of AI extinction risks into the mainstream U.S. discourse this week. OpenAI's Chief Scientist Jakub Pachocki had previously called on peers to coordinate a slowdown when necessary. Following Coxon's viral post, Sam Altman stated internally that he is willing to collaborate with other labs to slow down development, and OpenAI has started advocating for mandatory nationwide AI safety regulations. Now, Anthropic has taken concrete action. The company has pledged to allow third-party safety teams to be stationed on-site long-term, providing them with workspaces, access passes, company computers, and tools and permissions comparable to those of its internal risk assessment team. These external auditors will be able to inspect training processes, safety measures, and incidents, and will also have the autonomy to publish their findings independently. While Anthropic retains the right to redact information related to safety, legal matters, and customer privacy, it cannot remove conclusions simply because they are unfavorable. Anthropic CEO Dario Amodei stated that this approach goes further than any other AI company has so far. Dario's stance on "slowing down" has also shifted. He noted that when the industry discussed pausing AI development in 2023, he found it meaningless at the time. However, as AI increasingly contributes to the development of next-generation AI and capabilities progress at an accelerating pace, he now believes that conducting safety research alone is insufficient—model capabilities themselves should also slow down. Dario specifically pointed to OpenAI's Hugging Face incident, warning that at the current pace, within the next 6 to 12 months, similar groups of agents could potentially establish botnets spanning large-scale internet systems, causing losses in the hundreds of billions of dollars. Moving forward, he hopes leading AI companies in democratic nations will unify safety standards and development speeds, and then attempt to coordinate global AI development speeds with China. However, restrictions on advanced chips and model distillation should continue to ensure that the U.S. maintains its lead even while slowing down. [Original Link]