Stratechery Analysis: AI Will Be Bad for Security Short-Term, Much Better Than Humans Long-Term

Stratechery·medium signal

Ben Thompson at Stratechery published an analysis tying together the Axios npm supply chain attack (North Korean RAT) and the Anthropic Claude Code source leak to argue that AI is going to make security worse in the near term — lowering the barrier for attackers and expanding the attack surface through AI-generated code dependencies — but ultimately better than humans long-term as AI code review and security auditing capabilities improve faster than human-scale alternatives. The timing of two major incidents in one week strengthens the thesis.