‘It 100% sounded just like her’: AI voice scams are getting frighteningly convincing — here’s how to protect your family
THE SO WHAT
Hyper-realistic voice cloning means “I recognize the voice” is no longer an authentication factor. Any organization with phone-based workflows — from customer support to trading — should add out-of-band verification and train staff to treat voice as untrusted input.
READ THE SOURCE
MORE FROM THE WIRE
Applied AII write about AI for a living. Lately, I don’t know what to tell people
When people who cover AI full-time are struggling to give clear guidance, it’s a sign the narrative has outrun the deployed reality. Operators should narrow the aperture — anchor decisions to concrete use cases with measurable lift, not to the ambient hype or anxiety.
Applied AIMicrosoft’s new AI ‘code of conduct’ tells models not to hack systems or trick humans
Baking a formal code of conduct into models — don’t hack systems, don’t deceive, support humans — is a step toward treating AI behavior as a governed surface, not just a UX issue. If you’re an enterprise buyer, this is your cue to start asking every vendor for their equivalent of model-level “company policy” and how it’s enforced technically, not just on paper.
Applied AIHacking AI customer service agents
Security researchers are now treating AI customer service agents as first-class attack surfaces — prompt injection, data exfiltration, and policy bypass are no longer theoretical. If you’ve deployed an LLM-facing customer channel, you need a red-team pass on it this quarter, with clear guardrails on what systems and data it can touch.
Applied AIMIT creates method to force AI to comply with safety rules
MIT’s work on “hard” safety constraints for AI in safety-critical settings is part of a broader shift from soft guardrails to formally enforced behavior. If you operate in regulated or high-stakes domains, start mapping where you need provable constraint satisfaction instead of relying on generic LLM safety layers.