Phishing in the age of AI: why the mail looks real now

Robot hand reaching towards a network of connected points, illustrating AI

The badly written mail from a foreign prince is gone. Today’s phishing mail is in perfect Danish, refers to a real supplier and arrives at the exact moment you expect an invoice. AI made that easy.

The numbers

ENISA, the EU cyber security agency, analysed close to 4,900 incidents between July 2024 and June 2025. Phishing was the way in for around 60% of them. ENISA also reports that by early 2025, AI-supported phishing made up more than 80% of observed social engineering activity worldwide.

What AI changes

  • Perfect language. Spelling mistakes and odd phrasing no longer give the mail away.
  • Personal details. Names, titles and projects are taken from LinkedIn and company websites in seconds.
  • Volume. Thousands of tailored mails cost the same as one.
  • New channels. Text messages, Teams messages, phone calls and even cloned voices are used alongside email.

Five signs to look for

  1. Urgency: pay today, log in now, or the account is closed.
  2. A change to something you know, such as new bank details for a known supplier.
  3. A link or a login page that does not match the company’s real address.
  4. A request to keep something quiet or to skip the normal process.
  5. An attachment or a QR code you did not expect.

Five controls that stop most of it

  1. Two-factor login everywhere. A stolen password alone is then not enough.
  2. Call back on a known number before any change to payment details. Never use the number in the mail.
  3. Filter and label external mail, so people can see when a mail comes from outside.
  4. Short, regular training with real examples, instead of one long course a year.
  5. Make it easy to report. A report button and a quick thank you get more mails reported, and faster.

How we work at CC-Interactive

We use two-factor login on every system and give our people regular security and phishing training, because most attacks start with an email. Read more about the security measures in our work.

Want a system where approvals and payment changes always follow a fixed, logged flow? We build that. Book a meeting.

Source: ENISA Threat Landscape 2025.

See SmartBoard with your own data

Book a 30-minute demo. We look at your current reporting and show where robots and AI can take over, often starting with a small pilot that pays for itself.