AI CEOs Agree on Slowing Pace Amid Strategic Tensions
Top AI leaders support slower development, but global competition complicates binding safety rules.
Key points
- Anthropic and OpenAI CEOs committed to allowing independent evaluators full access to their systems.
- Over 1,300 AI employees have urged Washington to support an international slowdown.
- Strategic competition between the US and China creates a barrier to binding safety rules.
On September 12, leaders of major AI companies called for slowing the pace of development. This move follows growing concerns about safety risks in advanced models.
Anthropic and OpenAI committed to allowing independent evaluators full access to their systems. These pledges aim to address fears that current development speed outpaces safety controls.
Leaders Agree on Safety Measures
Dario Amodei of Anthropic proposed easing the pace of model improvements. He also suggested democratic coordination on shared safety limits globally. Sam Altman of OpenAI quickly agreed with this approach.
Elon Musk also supported the idea, stating that Amodei was right. This convergence is notable given previous disagreements within the industry. It reflects a shared recognition of potential catastrophic risks.
Strategic Competition Complicates Cooperation
However, structural incentives prevent a simple slowdown. The United States and China fear that slowing down will hand a strategic advantage to the other. This dynamic creates a nested collective-action problem.
More than 1,300 employees at leading AI companies have called for an international slowdown. They argue that the current race ignores serious safety concerns. Yet, government officials often use the global race as an excuse to avoid binding domestic regulation.
Trust Over Coercion Needed
Experts describe this situation as a “stag hunt” game. In this scenario, cooperation yields the best outcome for everyone. Defecting to chase a smaller, individual gain risks losing the greater shared benefit.
Noema Magazine notes that the solution requires assurance mechanisms. Players must believe others will keep their commitments. Verification and transparency are key to building this trust among rivals.
