Trump Opposes Amodei on AI Pace

Dario Amodei's call to slow down AI development reignites tensions with Donald Trump, who dismisses the idea of new safeguards. In the background, Anthropic is just emerging from a legal battle with the administration and operates without federal AI laws. Voices from the industry, from Elon Musk to a resigning researcher at Anthropic, are joining the debate.

Anthropic Between Public Disputes and Federal Ambiguity

Anthropic, which is preparing for a historic IPO, has had a contentious relationship with the Trump administration. In February, the Pentagon blacklisted the company amid a contractual dispute. Anthropic then initiated legal proceedings to contest these measures as well as a separate order prohibiting federal agencies from using its services. Last month, federal judge Rita Lin ruled this action illegal and unfounded. More recently, Commerce Secretary Howard Lutnick stated that Anthropic was back on good terms with the administration, expressing trust in the company and indicating that it had met expectations. This journey occurs in a context where cooperation on AI regulation remains challenging, even in the United States: the most extensive regulations currently fall under state jurisdiction, and Congress has not passed federal legislation despite numerous proposals.

Trump Rejects Safeguards, Musk Supports Amodei

Following the publication of an essay by Dario Amodei calling for a slowdown in global AI development, Donald Trump publicly criticized the Anthropic leader. On Truth Social, he claimed that his administration had prevented AI actors, including Amodei, from doing bad or potentially bad things, and that it would continue to do so. Trump reiterated his stance that the United States does not need additional regulations in AI, asserting that the only necessary safeguard is a strong and intelligent president, of which the country has an abundance. He also claimed that a harmful conspiracy was targeting AI and data centers, and that China was the only one benefiting from it, issuing a warning to "conspiracy theorists" and "traitors." In contrast, Elon Musk, CEO of SpaceX, publicly supported Dario Amodei by writing on X: "Dario is right."

Amodei's Three-Phase Plan to Slow Down AI

Dario Amodei explained on Saturday that recent developments, including OpenAI models deemed out of control, convinced him of the need for AI labs, including Anthropic, to slow down development. He advocates for reducing the pace of capability improvements in models and using the time gained thoughtfully. His plan unfolds in three phases, the first of which is already in place at Anthropic: external evaluators can access the company's systems under conditions comparable to those of employees, to ensure compliance with safety requirements, incident reporting, and model alignment. The subsequent phases would involve collaboration among democratic nations, and then on an international scale.

An Internal Alert Revives AI Security Concerns

The debate over AI security has been reignited by the viral resignation of Anthropic researcher Jacob Coxon. He claims that neither Anthropic nor OpenAI are acting responsibly. Last week, he wrote that they are rushing toward a self-improving superintelligence and playing with lives.