The scam got personal
Old-school phishing was full of spelling mistakes and looked like a scam. With generative AI, criminals produce messages in seconds that are tailored to the victim's job, the company's tone and data gathered from several platforms. ESET analysis points to a sharp rise in AI-generated campaigns.
"A call from a director asking for an urgent transfer could be a synthetic voice."
Scams also use audio and video deepfakes in corporate fraud: a 'call from the director' asking for an urgent transfer may be fake.
Why technology alone won't fix it
Email filters help, but something always gets through. When a message looks legitimate, what protects you is the habit of doubting and checking: confirming a payment request through another channel, checking senders and having an easy way to report anything suspicious.
That's why awareness can't be a once-a-year talk anymore. It has to be ongoing practice, with realistic simulations. Teams that practice recognize the pressure tactics and don't take the bait.
Defense on two fronts
On the technical side: a dedicated anti-phishing filter, sender authentication (SPF, DKIM and DMARC) and identity protection with MFA. On the human side: regular training, simulations and the habit of double-checking any unusual request.
AI is used on defense too: some tools analyze behavior and block anything out of the ordinary. They work best when people are ready for whatever slips past the filter.
Sources: WeLiveSecurity (ESET) and industry surveys on the use of AI in attacks (2025 and 2026).