**[Altman and Dario Call for Slowing Down AI, Trump Tech Advisor Sacks: If You're Really Worried, Stop Yourself, Don't Drag the Whole Industry]** Beating AI News Flash: Anthropic CEO Dario Amodei published a lengthy article calling for a slowdown in the advancement of cutting-edge AI capabilities. He believes that AI is increasingly involved in the development of next-generation AI, coupled with several recent incidents of agents losing control, and that safety research is struggling to keep up with the pace of capability growth. Dario proposed a three-step plan: first, allow third-party evaluators long-term access to AI companies with permissions similar to those of employees; second, have frontier labs in democratic countries coordinate safety standards and, if necessary, collectively slow down capability growth; and finally, promote international coordination. Anthropic has already committed to implementing the first step. Sam Altman subsequently expressed public support, stating that OpenAI has also been discussing "pacing the frontier" in recent weeks and announced that OpenAI would introduce similar long-term independent evaluators. Former White House AI and cryptocurrency chief, now co-chair of the President's Council of Advisors on Science and Technology, David Sacks, responded: "If you truly think internal models are dangerous, then stop yourself." He argued that OpenAI and Anthropic already dominate frontier AI and are fully capable of slowing down on their own without involving other companies. Sacks is particularly opposed to competitors collectively agreeing to "slow down together." Dario himself admitted that such coordination might require antitrust exemptions. Sacks believes that if government approval is needed to allow the largest AI companies to sit down and collectively limit speed, it would resemble a monopolistic alliance. He also criticized METR, questioning whether it is suitable to act as an independent referee for the entire industry. METR has long conducted external evaluations for companies like Anthropic and OpenAI, and recently, Anthropic safety researcher Joe Benton left to join METR. However, METR has explicitly stated that it does not accept funding from these frontier AI companies and only accepts model access and free tokens. Sacks further argued that slowing down aligns with the commercial interests of OpenAI and Anthropic. The easier it is for models to lose control, the greater the legal liability; customers are also more inclined to purchase stable and controllable AI. As for global unified deceleration, he predicts that not all countries will be willing to join. [Original Article Link]