AI Scams and Deepfake Fraud. When the Voice on the Phone Is Not Your Boss
Criminals can now copy a voice, or fake a video call, convincingly enough to fool careful people. The defence is a process, not better eyesight.
Fraud has always relied on pretending to be someone you trust. What has changed is the tooling. Generative AI makes it much easier to write polished, error free phishing emails, and security agencies and researchers have warned that it can also be used to clone voices and create convincing fake video of real people.
The typical target is not a clever technical system. It is a person under time pressure. A finance team member receives a call or message that appears to come from a director asking for an urgent payment, a change of supplier bank details, or gift cards. The story is plausible, the voice sounds right, and the request is awkward to question. That is exactly what the criminal is counting on.
You cannot reliably spot a good fake by looking or listening, so build checks that do not depend on it. Any request to move money or change bank details should be confirmed through a different channel, using a phone number you already hold, not one supplied in the message. Agree that two people must approve payments above a set amount, and that no one will be criticised for pausing to check, even if the request seems to come from the managing director.
Some firms also agree a private code word for urgent requests. It is a simple idea and it can help, as long as it is kept offline and not widely shared. Be careful what you post publicly, because recordings of voices and detailed organisation charts give criminals raw material.
Technical controls still matter. Strong email protection and multifactor authentication reduce the chance of the criminal getting into a real mailbox in the first place. Training should include these scenarios in plain language. If you would like help tightening your payment checks or staff awareness, talk to Dragonfly IT.
