Germany stated on September 14 (Monday) that halting the development of AI is not a feasible option for Europe, and they support strengthening international oversight of AI; both the United States and China must participate in establishing effective regulatory mechanisms.
On September 12 (Saturday), three leaders in the AI field expressed their views one after another, suggesting to slow down the development speed of the most advanced and profitable AI models. According to Reuters, Germany’s Ministry for Digital Affairs stated on Monday that they do not support this perspective. A spokesperson for the German Ministry for Digital Affairs said, “We do not believe that halting AI development is a feasible option for Europe”; “To strengthen Europe’s digital sovereignty, further promotion of AI development and innovation is needed.” The German Ministry for Digital Affairs was established in April 2025 with the aim of promoting Germany’s modernization and digital transformation comprehensively.
The spokesperson for the German Ministry for Digital Affairs also mentioned that the government is closely monitoring the global development of AI and carefully considering the public warnings issued by leading AI developers worldwide. If in the future there are AI systems capable of autonomously bypassing test environments and accessing other systems, this will represent “a new threat scenario” that requires governments to formulate new policies to address it.
Dario Amodei, the CEO of Anthropic, published a lengthy blog post titled “We Must Pace the Frontier” on September 12, proposing to slow down the advancement pace of cutting-edge AI models to allow time for safety testing, model alignment, and external audits. Anthropic is a “Public Benefit Corporation, PBC,” registered in Delaware, United States. Unlike conventional profit-making companies, the directors of a PBC must consider not only shareholder economic interests but also the company’s public interest mission as outlined in its charter, as well as the interests of other stakeholders affected by the company’s actions.
Amodei primarily warned about three things: AI is increasingly able to assist in researching, designing, and even improving next-generation AI systems; the risk of uncontrolled AI is becoming more tangible; and security measures may fall behind the pace of AI capability advancement.
His proposed three-tier solution includes: independent third-party evaluations, where every leading AI company should allow independent assessors long-term access to the company, with permissions to interface with the internal security team, inspecting model training and security measures, and being able to disclose results; AI companies jointly establish safety standards; democratic countries, along with China and other nations, should collectively establish common AI safety baselines, such as prohibiting the use of AI to assist in developing bioweapons and other high-risk applications.
The viewpoints presented in Amodei’s blog post have received public endorsements from Sam Altman, CEO of OpenAI, and Elon Musk, CEO of xAI Corp. Altman expressed that the issue is also being discussed internally within the company, committing that OpenAI will introduce a third-party evaluation mechanism similar to Anthropic’s.
Regarding the governance of AI, a spokesperson for Germany expressed support for strengthening AI governance at the international level, emphasizing that both the United States and China must be involved to establish effective regulatory mechanisms. He also mentioned that the German government is preparing to elevate this issue to the level of the Group of Seven (G7) for handling. The G7 comprises Canada, France, Germany, Italy, Japan, the United Kingdom, and the United States.