Phishing in minutes, not weeks
A fake bank text that almost got me, a deepfake video call that cost a company millions, and why the answer to AI-powered scams is to get sharper, not to stop AI.
A few weeks ago, a text message landed in the same thread as my real bank's messages. Same sender name. Same tone. It said a large payment was waiting for my approval, and if it wasn't me, I should tap the link to cancel it right away.
My thumb was already moving. I'd had a long day, and "cancel a payment I didn't make" sounded like exactly the right thing to do.
Then I noticed the link. The bank's name was there, but spelled with one extra letter.
I work in tech. I build software for a living. I know all the warning signs. And I still almost fell for it.
Scams used to take time
Not long ago, a convincing phishing attack took real effort. Someone had to copy a website by hand, write the messages, fix the broken English, and hope nobody looked too closely.
Now AI does most of that in minutes. A fake login page that looks pixel-perfect. Emails written in flawless, friendly language, in any language you speak. Messages personalized with details from your public profiles. What used to take days or weeks now takes a coffee break.
And it doesn't stop at text. Voices can be cloned from a short clip. Faces can be faked on a live video call. In early 2024, an employee at the engineering firm Arup in Hong Kong joined a video call with what looked like the company's CFO and other colleagues. They were all deepfakes. He was asked to make transfers, and he sent about HK$200 million to the criminals.
Seeing used to be believing. It isn't anymore.
Every tool has a dark side
This is the dark side of AI. But I don't think the answer is to fear it or stop it.
Every powerful tool in history has been used both ways. The same knife cooks dinner. The same internet connects families and spreads scams. AI is no different. The same technology that writes the fake email also powers the filters that catch it.
So the better path is to embrace AI, and get sharper alongside it.
A few habits that still work
- Slow down when it feels urgent. Urgency is the scammer's favorite tool. A real problem can wait two minutes.
- Check the address, not the name. Logos, names, and writing style are easy to fake. The actual link or email address is harder.
- Verify through a second channel. Got a strange request from your boss? Call him on the number you already have, not the one in the message.
- Never share one-time codes. No real bank, colleague, or support team needs them.
- Agree on a code word with your family or your finance team for any urgent money request.
The higher you are, the bigger the target
If you lead a company or a team, you're not just a target. You're a key. Your voice, your face, and your approval can open doors that matter.
That means more care, not less. Never approve a big transfer from a call or a chat alone. Set rules that need two people to agree. Talk to your team about these scams openly, so nobody feels silly asking "Is this really you?"
With great power comes great responsibility, as Spider-Man's Uncle Ben famously put it. In the AI era, part of that responsibility is simply being harder to fool.
Stay curious about AI. Stay careful with it too.
Keep reading
More from the notebook.
Nobody deletes an app because of its framework
A friend asked whether to build native or use Flutter. Why AI may bring big companies back to native apps, and why a new product should still ship first.
Praying to the servers
An index that froze the orders table, a shrine for praying while we waited, and why the best server prayer is the one nobody has to say.