AI-assisted scams
Attackers now use AI tools too. The tricks are the same as always; AI just makes them faster, cheaper, and more convincing. Knowing what AI can fake helps you stay a step ahead.
What AI changes
- Flawless writing. The spelling mistakes that once gave phishing away are gone. A polished, on-brand email is no longer proof it is genuine.
- Voice cloning. A short clip of someone's voice is enough to fake a phone call. A panicked "it's me, I need you to pay this now" may not be them at all.
- Deepfake video. Faces and video calls can be faked well enough to fool a quick glance, including in a live meeting.
- Personalised at scale. AI can tailor thousands of messages using details scraped from the web, so each one feels written just for you.
The defence has not changed
Because the content can be faked, trust the process, not the polish:
- Verify urgent or unusual requests out of band, through a channel you already trust. This stops a cloned voice or a deepfake cold.
- Agree a code word for high-stakes requests among family or finance teams, so you can confirm a real person.
- Be careful what you post. Voice clips, photos, and personal details are raw material for a convincing fake.
- Use only approved AI tools for work data. Pasting sensitive information into a public AI tool sends it to someone else's servers.
If a message is surprising, emotional, and in a hurry, that is exactly when to slow down and confirm.