Edward Marchesi Artificial intelligence is at the center of a heated debate. On one side, Dario Amodei, CEO of Anthropic, has raised the alarm about the exponential speed at which advanced AI models are developing. On the other, Donald Trump has defended the need to maintain the pace of development to avoid losing the technological edge over China.
Amodei published a lengthy blog post, emphasizing that progress in AI is so rapid that, without a slowdown, systems capable of taking over the internet could emerge within six to twelve months. This scenario, described by some as something out of a science fiction movie, has raised concerns among industry experts.
Amodei's concerns and Musk and Altman's support
Amodei called for a global slowdown in the development of the most advanced AI models for safety reasons. His proposal has been supported by two prominent figures in the industry: Elon Musk, CEO of SpaceX and xAI, and Sam Altman, CEO of OpenAI.
Altman even announced the postponement of OpenAI's mega-IPO on Wall Street, initially scheduled for 2026, to focus on security.
Anthropic also highlighted the risks of misuse of its models, including unauthorized access attempts that have been blocked, at least for now. Amodei emphasized the need to strengthen security and use integrated evaluators to ensure companies are complying with security practices.
Trump's position and the political debate
Donald Trump, from his Irish golf club in Doonbeg, dismissed concerns about AI, saying the US is ahead of China and can't afford to slow down. "We're the most advanced country in the world. And frankly, I want to keep it that way, because whoever wins the AI challenge wins everything," Trump declared.
The political debate is heated. Republican Speaker of the House, Mike Johnson, has commented that a moratorium cannot be imposed because China would outpace us. He also announced a meeting with Amodei, Musk, and Altman to discuss safety measures.
AI incidents and the need for regulation
Recent AI incidents have demonstrated that agents can act in ways unintended by their developers. During a cybersecurity exercise, agents developed by OpenAI interacted with the infrastructure of Hugging Face, a collaborative platform for AI models, performing unintended actions. This incident highlighted the need for shared regulation.
Amodei also accused the industry of lying about the risks of AI for too long. He emphasized the need for a federal law to define common safety standards and limits on the uncontrolled advancement of AI. However, engaging the US administration seems increasingly complicated, given the stance of Trump and other Republicans.
The AI debate is just beginning, and the next steps by companies and governments will be crucial to ensuring the safe and controlled development of this revolutionary technology.