Why an AI slowdown needs Beijing to feel the risk

# What happened this week

# Why leaders are calling for a slowdown Dario Amodei argued that capability growth needs to slow so risk-reduction work can catch up. He cited recent troubling events: an incident where a swarm of AI agents attacked a company (the "Hugging Face attack") and the growing prospect of AI systems that can improve themselves without human help (recursive self-improvement). These developments convinced several leaders that rapid capability gains are outpacing available safety measures.

# The public and political ripple effects

# The core obstacle: China's incentives Noah Smith's piece argues that asking major US AI firms to slow down is only half the problem. If the goal is a meaningful global pause or coordinated pacing, China must be part of any durable agreement. Beijing has made AI a national priority. Chinese policymakers weigh several concrete factors when deciding how to respond:

• Strategic advantage: A voluntary US slowdown could be framed in China as an attempt to preserve American dominance. Beijing will be wary of any action that might lock in a foreign lead.
• Industrial policy and jobs: China treats AI as an economic and social modernization project. Slowing development can conflict with plans to boost productivity and manage worker displacement.
• Domestic stability: If AI threatens employment or social cohesion, Chinese leaders might prefer controlled deployment rather than an open-ended international pause.

These incentives mean China will not be persuaded by US-led moral arguments alone. For a slowdown to take hold, Beijing would need evidence the risks are real and immediate and assurances that slowing won't result in a strategic or economic loss.

# What might convince China The story suggests several practical levers that could matter to Beijing:

• Shared, verifiable safety standards with balanced enforcement and independent evaluators that include Chinese participation.

• Demonstrable incidents showing capability failures or misuse at scales that threaten social stability or national security.

# Bottom line US AI leaders can advocate pacing, and recent events have given those calls unusual political traction. But a unilateral or Western-only slowdown is unlikely to be effective. Any meaningful pause requires convincing China that the risks are immediate and that compliance is in its national interest. That requires concrete, reciprocal mechanisms and proof that slowing will not convert into long-term geopolitical leverage for one side.