Deepfake and voice-fraud intelligence from the people building the detection. Real attacks, new generation models, what defends against them, and where the regulation is heading.
AI voice cloning makes familiarity a weak control, so regulated firms need evidence that staff can follow independent verification under pressure.
Read the post →A convincing face and familiar voice should never be able to bypass callback, approval, separation-of-duties and escalation controls.
Read the post →A reported voice-cloning fraud case is a reminder that resilience depends on workflow design, escalation and evidence, not on asking staff to judge whether a voice sounds genuine.
Read the post →Office background noise hid a cloned voice from our own detector. The cause was how detectors average a call, not the model. Here is the test, the fix, and what to check in yours.
Read the post →The evidence is moving from theoretical threat to board-level risk, but many firms still test training and identity checks rather than the phone workflows attackers are targeting.
Read the post →Two stories, eleven days apart, that belong together, and the enforcement gap between them.
Read the post →An open-source app clones a voice in 646 languages from three seconds of audio, locally and free. The threat model just changed.
Read the post →