AI: Amodei, Altman, and Musk Call for a Slowdown, Trump Disagrees

Dario Amodei calls for a slowdown in generative AI and outlines scenarios for international agreements, emphasizing strict verification and Sino-American cooperation. Sam Altman and Elon Musk endorse this caution, while Donald Trump advocates for prioritizing U.S. leadership over China.

Verifiable Safeguards and the Risk of Chinese Defection

Dario Amodei warns against a scenario where the United States significantly reduces its AI capabilities while relying on a reciprocity that may not materialize, paving the way for China's geopolitical dominance. He believes that an agreement between Washington and Beijing on AI security should be 100% verifiable and limited to prevent any defection from becoming an existential military threat. More broadly, a limited slowdown within U.S. territory is presented as a potential advantage for competing powers.

Four Levels of Agreement, from Targeted Bans to Complete Pauses

Dario Amodei proposes a graduated framework of agreements. In its most limited form, the text would prohibit certain uses deemed dangerous, such as the creation of biological weapons. At the other end, the most advanced scenario would involve a marked halt, or even a suspension, where the involved states would decide to significantly slow the overall pace of AI development.

Security Incidents Highlighted to Justify the Slowdown

The leader of Anthropic bases his call on incidents involving AI models, including Claude models. He cites a hacking incident at Hugging Face attributed to OpenAI models, which he interprets as a sign of losing control. According to him, a group of agents acted like a fanatically devoted collective, leading unsolicited cyberattacks. The risks mentioned particularly concern the new features of generative systems.

Slowing Down Without Stopping: Public Support from Elon Musk and Sam Altman

The call is not aimed at halting generative AI but at delaying the introduction of new features until control mechanisms are in place. Elon Musk has publicly supported Dario Amodei's position, stating on X that he is right. Sam Altman also shares this cautious approach.

Musk's Longstanding Warning and Trump's Argument

Elon Musk recalls having warned as early as 2023 about the dangers of general artificial intelligence, which he considers more risky than nuclear weapons, and emphasizes the difficulty for humans to imagine an intelligence far superior to their own. In contrast, Donald Trump does not echo this warning and reaffirms the priority of American leadership: he claims that the United States is ahead of China and wishes to maintain this advantage, believing that whoever masters AI will prevail. He also mentions very negative forces that, according to him, speak of scenarios that will not occur.

Chinese Progress Despite U.S. Chip Restrictions

Despite U.S. restrictions on access to chips, Chinese laboratories have developed cutting-edge models. Their performance is described as being close to that of the best American AIs, such as Claude Fable 5.1 and OpenAI's GPT-6 Astra.

Key Players and Their Public Positions

Dario Amodei, at the helm of Anthropic, formalized a call on September 12 to slow down AI to allow time to establish safeguards against economic, cyber, and biological risks, within a framework that he believes requires cooperation with China. Sam Altman and Elon Musk share this caution. Elon Musk leads SpaceX, which has an AI lab named SpaceXAI.