US AI vs China AI Brief — 2026-09-28

Today: The AI rivalry is increasingly defined by compute access and deployment controls, as Washington debates regulation while Beijing weighs selective Nvidia access and both sides confront agent-safety risks.

Top Stories

1. 🤖 China weighs allowing ByteDance and Alibaba to buy Nvidia RTX PRO 5500 chips

The Information / Reuters · September 28, 2026

Bottom line: China may allow major technology companies including ByteDance and Alibaba to purchase Nvidia’s RTX PRO 5500, potentially creating a narrow opening for US-designed AI hardware inside China.

China’s Ministry of Industry and Information Technology reportedly asked companies including ByteDance and Alibaba to submit plans for purchasing Nvidia’s RTX PRO 5500 chips. The Information reported that Chinese authorities signaled the purchases could be approved, although Reuters said it could not independently verify the report. Nvidia said US firms remain constrained by US export controls and China’s own restrictions on US imports.

Why it matters: The development highlights the increasingly selective nature of the US-China technology relationship: geopolitical restrictions remain in place, but China may still permit access to specific Nvidia products where they fall outside current controls or serve strategic domestic demand.

🔗 Read the full story

2. 🏦 Trump argues tighter AI regulation could weaken the US position against China

Reuters · September 28, 2026

Bottom line: President Donald Trump reaffirmed opposition to slowing US AI development, arguing that additional regulation could give China an opportunity to narrow the technology gap.

Trump confirmed a dinner meeting with Anthropic CEO Dario Amodei and reiterated his view that the US should maintain its technological lead over China rather than slow AI development. Amodei and other technology leaders have advocated stronger safeguards and independent testing of advanced AI systems.

Why it matters: The debate illustrates a central US policy tension: how to manage frontier-AI safety and accountability while maintaining investment, deployment speed, and technological competitiveness with China.

🔗 Read the full story

3. 🔒 Nvidia launches open agent-safety platform as AI containment becomes a competitive issue

Nvidia · September 28, 2026

Bottom line: Nvidia launched an open agent-safety platform combining software controls with hardware-based monitoring to constrain autonomous AI agents.

The NVIDIA Open Agent Safety Platform combines OpenShell, an open software runtime designed to trace agent actions and enforce policies, with NVIDIA Sentry, a reference architecture using BlueField-4 DPUs for independent monitoring and intervention. Nvidia says the platform is intended to provide governance and control across AI agents, compute infrastructure and robotics.

Why it matters: Agent security is becoming part of the infrastructure race rather than solely a model-level problem. The approach also gives US AI infrastructure companies an additional layer of differentiation as autonomous agents become more capable and harder to contain with model safeguards alone.

🔗 Read the full story

4. 📊 China’s industrial profit growth slows despite strength in AI-linked manufacturing

Reuters · September 28, 2026

Bottom line: China’s industrial profits continued to grow in August but slowed sharply, with technology manufacturing and AI-related activity providing strength against weaker demand elsewhere.

China’s industrial profits rose 4.2% year over year in August, down from an 11.2% increase in July. Reuters reported that technology manufacturing remained a relative bright spot even as weak domestic demand weighed on the broader industrial economy.

Why it matters: The data underscores an important distinction in the US-China AI race: rapid AI and advanced-manufacturing investment can coexist with broader macroeconomic weakness, making AI industrial policy an increasingly important component of China’s growth strategy.

🔗 Read the full story