AI Leaders Urge Slowdown as Trump Rejects Losing Ground to China
Frontier AI systems are nearing capabilities that can autonomously handle complex tasks, while safety testing, alignment research and regulation risk lagging behind. That tension has put rival developers Anthropic, OpenAI and Elon Musk’s xAI in rare agreement that the industry should pace frontier-model advances. Their stance matters because a unilateral slowdown could reduce catastrophic risks but also surrender commercial and strategic ground, exposing the conflict between AI governance and Washington’s drive to stay ahead of China.
On Sept. 12, Anthropic Chief Executive Dario Amodei warned that, without a slowdown, AI could within six to 12 months direct swarms of agents capable of taking over the internet, potentially causing hundreds of billions of dollars in damage. He said gaining one or two years could materially advance alignment work. OpenAI CEO Sam Altman and xAI founder Elon Musk backed the call the same day. President Donald Trump rejected slowing the race on Sept. 13 while speaking in Ireland, saying the United States leads China and “whoever wins AI wins,” though he acknowledged a need for some guardrails.
All Coverage
1 original reportsThe Backstory
The history behind this eventTrump Rejects AI Slowdown, Puts China Race First
Rapid advances in generative artificial intelligence have intensified debate over whether developers should slow work on systems that could surpass human capabilities. Anthropic Chief Executive Dario Amodei and other industry figures, including OpenAI CEO Sam Altman, have raised concerns about superintelligence and called for stronger safeguards, putting technological risk in tension with Washington’s drive to stay ahead of China.
President Donald Trump said in Doonbeg, Ireland, on Sept. 14, 2026, that the United States led China in AI and must preserve that advantage, declaring that whoever wins AI “wins.” He dismissed slowdown advocates as negative forces discussing scenarios he said would not happen, while acknowledging that guardrails could be adopted. The dispute sharpened after former Anthropic and OpenAI researcher Jacob Coxon resigned and warned that AI could threaten humanity before the end of the decade.
OpenAI Moves Frontier AI Safety Checks Earlier as Trump Opposes Slowdown
Rapid advances in frontier AI have raised concerns that model capabilities could outpace safeguards designed to keep systems aligned with human intent, particularly in cybersecurity, biological risk and autonomous behavior. Anthropic has pushed developers to build safety thresholds into the model-development process, an approach backed by OpenAI Chief Executive Sam Altman. The debate matters because leading laboratories are racing to train more powerful systems while governments weigh how to contain potentially severe risks without surrendering technological and economic leadership.
OpenAI said its latest approach would move safety evaluations for frontier AI to before model training begins, allowing risk controls and alignment measures to be considered ahead of capability gains. US President Donald Trump opposed slowing development over safety concerns, arguing that the United States must preserve its competitive advantage over China in AI. The report did not specify an effective date, quantitative safety thresholds or any financial commitment, leaving the timetable and enforcement details of OpenAI’s new measures unclear.
Subscribe to Mark Radar Weekly
Every Friday, the week's strongest signals in your inbox. Unsubscribe anytime.
If you search news on Google, you can set Mark Radar as a preferred source—our coverage will show up more often in your results. Set as preferred source on Google →