# Man and Machine: When the Scam Reaches Your AI

![Man and Machine: When the Scam Reaches Your AI](/writing/man-and-machine/when-the-scam-reaches-your-ai/assets/hero.jpg)

Imagine your AI assistant checking email when an invoice arrives asking you to pay a familiar supplier into a new bank account\. It could be genuine or a convincing fake\. Your assistant might help notice the changed details, but if it can also send money, one bad judgment could become a financial loss\.

AI is part of this problem and potentially part of the defense\. The FBI warned in 2024 that criminals use it to make fraudulent messages more believable\. That same year, Britain’s National Cyber Security Centre described AI’s potential to identify malicious emails\. Neither observation tells us that a personal assistant can reliably decide which requests are safe\.

I’m opening Defending Ourselves in the AI Age with two questions: what do we want AI to help us protect, and how much authority should we give it?

Start with one assistant\. Make a private list of the accounts and information it can reach, then note what it can read, change, send or spend\. Keep passwords and recovery codes out of the list\.

Circle anything it can do without asking you\. If reading a payment request and acting on it are both allowed, decide whether you want those permissions joined\.

A message can ask for money\. It doesn’t get to decide what your assistant is allowed to do\.

\-dg

\#ManAndMachine \#PersonalAI \#AISecurity

## Sources

- [Original LinkedIn post](https://www.linkedin.com/feed/update/urn:li:activity:7523111764335480833/)
