Summary

Cathy O'Neil argues that AI kill switches, a popular policy proposal, are fundamentally insufficient because they only address compliant domestic actors while leaving financial systems vulnerable to international criminal networks wielding AI agents. She points to payment giants like Mastercard and Visa embedding AI into shopping and transactions as evidence we are moving in the wrong direction. Her core prescription is twofold: structurally separate the financial system from AI agents, and establish criminal accountability for individuals who cause large-scale economic or physical harm through AI. Even these measures, she concedes, won't fully neutralize well-organized international crime rings.

Key Insight

Kill switches regulate the compliant, not the criminal — the real AI threat to finance comes from actors no domestic policy can reach, making structural separation of AI from financial systems the only durable defense.

Spicy Quotes (click to share)

• 7

  Even if we could kill responsible uses of AI by American companies, we won't be able to control for international criminal tech savvy rings using swarms of AI agents to attack our bank accounts.

• 6

  Ultimately we will have to separate our financial system from AI agents altogether.

• 8

  AI is good at math and good at spewing bullshit but it's really really good at hacking into systems.

• 6

  What we need to do is set up a system of accountability so that individuals get in trouble – and criminally charged – for large scale economic or physical harm via AI.

• 4

  Even if we do that, international crime rings will still be hard to prosecute.

Tone

critical, alarmed