THE REAL AI RISK IS INSIDE THE LABS

Summary from the TLDR Data newsletter

Frontier AI labs are more likely to cause the first serious AI incident than open models, since closed systems are vulnerable to insider leaks. Current open models also lack the domain depth to pose immediate risk in fields like biology, and restricting access helps malicious actors more than defenders.