Loading…
Topic
Export Controls
2 posts about Export Controls. Every summary links to the original.
- 2 posts
- 1 company
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
Newest first
Posts
Securing America's compute advantage: Anthropic’s position on the diffusion rule
Anthropic’s analysis of the Diffusion Rule, a January 2025 export control framework for advanced AI chips and model weights, warns that maintaining America’s compute advantage is essential for national security. The rule creates a three-tier system: Tier 1 allies face few restrictions, Tier 2 moderate limits, and Tier 3 adversarial nations strict controls. Anthropic recommends allowing Tier 2 countries with strong data center security to obtain more chips via government agreements, lowering the no-license threshold of 1,700 NVIDIA H100s, and boosting enforcement funding. The company notes that Chinese labs like DeepSeek, using chips obtained before controls, already face 2-4x higher energy costs and that by 2027 older-chip countries could pay ten times more for AI training. Anthropic warns against delaying implementation past May 15, 2025, as Chinese firms are stockpiling and a pause would weaken controls.
Thoughts on America’s AI Action Plan
The White House released its “Winning the Race: America’s AI Action Plan” to sustain U.S. AI leadership, and Anthropic responded supporting many recommendations while noting gaps. The plan aligns with Anthropic’s prior proposals on accelerating AI infrastructure, expanding federal adoption, and democratizing benefits through NAIRR and retraining programs. Anthropic strongly urges maintaining export controls on Nvidia H20 chips to China, arguing their memory bandwidth provides unique inference capabilities unmatched by domestic Chinese chips. The company also advocates for a single national standard on AI development transparency, requiring public reporting on safety testing and assessments for catastrophic risks like CBRN development. While the plan positions America for AI advancement, Anthropic believes these additional measures are essential for securing American leadership and ensuring safe development.
Related topics
All topics →Weekly digest
The best engineering reads, once a week.