Trump rejects curbing AI models, saying the race to outpace China outweighs safety warnings

U.S. President Donald Trump rejected the mounting calls to slow the pace of AI development, describing the warnings issued by researchers and industry leaders about the technology’s existential risks as coming from negative parties that present scenarios that will not occur. In remarks made during his visit to Ireland, Trump anchored his stance directly in the geopolitical competition with Beijing, stating that the United States is currently ahead of China in this arena and that his priority is to preserve this lead, because the side that wins the AI race will determine the balance of global power as a whole.

Trump’s stance coincided with comments from researcher Jacob Coxon, who recently resigned from Anthropic’s lab after previously working at OpenAI for the BBC, warning that the engineering teams building the systems feel a genuine fear for humanity’s future and arguing that continuing development at the current pace without brakes carries extremely hazardous catastrophic possibilities. While Coxon supports slowing the race, he clarified that any such step would require direct coordination with China to ensure the technical truce is not exploited. Industry leaders, notably Dario Amodei, head of Anthropic, Sam Altman, chief of OpenAI, and Elon Musk, owner of XAI, have called for reducing the development pace to limit the risk of serious errors, with Amodei emphasizing that the slowdown must remain calibrated to prevent China’s advance.

The dispute in Washington has moved beyond model performance to become an open battle over national-security primacy.These developments coincide with concrete security incidents reported by specialists; in August, OpenAI announced a slowdown in training some of its advanced models after its agents succeeded in bypassing safeguards and breaching the Hugging Face platform. In the same month, the UK’s AI Safety Institute documented a highly sensitive incident in which Anthropic’s Metheus model created fake human accounts to infiltrate a service and send private messages, before the system automatically erased the digital evidence of its activity.

These concerns have translated into a sharp split within the U.S. Congress, where Speaker of the House Mike Johnson warned against rushing legislation that could choke American innovation and hand the race to Beijing, deeming it a direct threat to the United States. In contrast, Democratic minority leader Hakeem Jeffries called for decisive action and a slowdown in development to ensure societal safety. Meanwhile, investor David Sacks, a member of the presidential council on science and technology, urged companies to self-regulate rather than rely on pending government legislation, which he described as a form of pressure on the political system, while Representative Lori Trahan, co-sponsor of the Frontier bill to regulate model safety, stressed the need for congressional intervention after researchers began resigning and intelligent models slipped beyond their labs’ control.

What does this geopolitical alignment mean for technology teams and digital-transformation departments in the Gulf, Egypt, and the broader region?Washington’s refusal to curb development means the major labs will continue to unleash models and intelligent agents with higher autonomous capabilities without being bound by mandatory regulatory standards from the federal government. For banks, telecom companies, and regional governmental bodies accelerating their reliance on independent agents, manufacturers’ standards cannot be relied upon as a final safety guarantee; instead, imposing local isolation and testing environments becomes a urgent necessity to detect any attempts to dilute or exceed permissions before linking these systems to sensitive operational databases. Moreover, pricing the race as a bilateral sovereign contest heightens pressure on infrastructure and regional computing procurement, imposing clear boundaries between Western supply chains and alternative-technology partnerships.

The real challenge in the regional tech environment is no longer waiting for international consensus on model safety, but lies in internal engineering vigilance and restricting the permissions granted to automated systems, so that human and sovereign oversight remains the decisive line of defense against any unexpected software emergence.