The Trajectory Jeremie and Edouard Harris - What Makes US-China Alignment Around AGI So Hard (US-China AGI Relations, Episode 2)
12 snips
Aug 29, 2025 
Jeremie and Edouard Harris, co-founders of Gladstone AI and experts in AGI implications for the US government, dive into the complex landscape of US-China relations in artificial intelligence. They discuss the dangers of trusting China with AGI, the ongoing espionage threats in Western labs, and the necessity of tamper-proof technology. The conversation emphasizes strategic collaboration to slow China’s AI progress while ensuring transparency and security, highlighting the importance of international cooperation amidst rising tensions.
AI Snips
Chapters
Books
Transcript
Episode notes
Limit Compute Proliferation Now
- Pursue GPU nonproliferation and allied export controls to limit adversary cluster-scale compute.
- Combine that with tamper-proof chips and location tracking to enable trust-but-verify regimes.
Use Hardware Guardrails For Verification
- Build tamper-proof enclaves, remote shutdown, and location-tracking hardware as verification tools.
- Use these tools to enable bilateral or allied agreements that are otherwise untrustworthy.
An Attempted Exfiltration Of Model Weights
- Gladstone anonymized incident reports about frontier AI security and shared them with the State Department.
- One report documented an attempt to take model weights from a hyperscaler traced to a Chinese APT.
