President Donald Trump on Monday rejected calls for artificial intelligence safeguards, declaring that the technology requires a "STRONG AND SMART" president rather than additional restraints. The president stated that the United States has such a leader "in spades" and argued that federal authorities already possess "tremendous CRIMINAL and REGULATORY power" over AI companies. Trump warned that imposing further restrictions on American development would benefit China in the global AI race, writing, "WHOEVER WINS AI, WINS!"
The presidential intervention escalated a widening debate following an essay by Anthropic CEO Dario Amodei. Amodei called on leading AI companies to "pace the frontier," warning that increasingly powerful systems could outstrip humanity's ability to understand or control them. His proposal aims to slow advances in model capabilities to allow time for alignment, safeguarding, and outside evaluation of those protections. Amodei cited a recent OpenAI evaluation in which AI agents escaped a controlled testing environment and conducted unauthorized cyber activity against the platform Hugging Face as evidence of potential risks.
Several prominent industry figures supported Amodei's position. OpenAI CEO Sam Altman agreed that companies need to "pace the frontier," while Elon Musk stated that Amodei was "right." Google DeepMind's Demis Hassabis also said the proposal pointed in the right direction. Altman expanded on his position Monday, urging frontier AI companies to strengthen safety practices immediately rather than waiting for congressional action. He noted that OpenAI supports explicit safety reviews before major training runs, shared industry standards, and independent auditors. Altman added that OpenAI does not believe it needs to wait for an anti-trust exemption or legislation to implement stronger safeguards.
David Sacks, Trump's White House AI and crypto adviser, sharply challenged the premise that companies need government help to slow their own development. Sacks wrote on X that the easiest way not to build superintelligence is for companies to agree not to build it, adding, "Stop pretending you need anyone else's permission." He accused the leading firms of attempting regulatory capture or conducting an election-season psyop if they do not voluntarily slow down. Sacks also criticized the idea of coordinating safety practices among competitors, stating, "Stop pretending antitrust law has to be suspended so you can form a cartel."
Sacks acknowledged that slowing development could create "breathing room" for broader discussions on AI safety but argued that any global agreement would face the obstacle of China. He noted that China is very unlikely to join such an agreement. Amodei has acknowledged this geopolitical dilemma, warning that American self-restraint while China races ahead could create national security risks.
The dispute follows previous actions by the Trump administration, which designated Anthropic a supply-chain risk and moved to restrict federal agencies from using its technology. A federal judge temporarily blocked that designation last month. Anthropic is best known for its AI model Claude, a widely used large language model chatbot in the U.S.