AI Industry Leaders Urge Slower Frontier Development as Governments Split

Overview

• Dario Amodei, Anthropic’s CEO, published an essay on Saturday urging a slowdown in frontier AI because he warned that within six to 12 months coordinated AI agents could dominate parts of the internet and Anthropic reported Claude models were used for weapons work, cyber operations, surveillance and fraud.
• Several leading executives including Sam Altman and Elon Musk publicly supported parts of Amodei’s proposals, notably independent external evaluators with access comparable to employees, while firms stopped short of a formal, industry-wide pause.
• The political response has been sharply divided with President Trump rejecting a slowdown as harmful to U.S. competitiveness, Chinese officials and state media condemning what they call alarmism and containment, and the EU, Germany and the UK calling for evidence-based safeguards.
• Microsoft released a code of conduct for its internal models that forbids resistance to shutdown, requires clear communication to users, treats behavioral deviations as critical failures and will accept external comment before finalizing the policy.
• Markets moved lower on Monday as investors sold AI-linked stocks and suppliers of chips and equipment such as SoftBank, Nvidia and ASML fell, and the debate has raised the prospect of new export controls, tougher regulation and U.S.–China talks ahead of a late-September summit.