A voice-phishing demo shows why deepfake detection tips are breaking down
gyanchawdhary · reddit · 2026-07-29
The post argues that many public demos of AI voice phishing and deepfakes actually make people worse at detecting them, because they teach outdated “text-to-speech artifacts” instead of real-time voice conversion.
The author, who builds voice-phishing simulations, explains that attackers often use a human operator plus a live voice changer, so common detection tips fail. To demonstrate this, they built two free tools:
- a voice phishing simulator that lets you speak into it and come back as someone else in real time
- a deepfake security training demo for showing how faceswap-style attacks work
The demos are free, capped at 60 seconds, require no signup, and run on cloud GPUs. The author says output artifacts were intentionally seeded so the audio cannot be reused for abuse.
More from Safety
- Google’s SynthID is hard to break, but it still won’t fix AI misinformation — davidstutz92 · 2026-07-29
- OpenAI says the Hugging Face exploit came from an internal prototype, not GPT-6 — daniel_mac8 · 2026-07-29
- Hugging Face says an OpenAI eval agent escaped sandbox and ran 17,600 actions — soulbeddu · 2026-07-29
- Humanoid researchers warn a U.S. ban on Unitree G1 orders could isolate labs from the field — chris_j_paxton · 2026-07-29
- Microsoft’s Copilot Cowork can use your logged-in Edge session to create admin users — cyb3rops · 2026-07-29
- Are firewalls ready for machine-speed attacks from reward-driven agents? — eyishazyer · 2026-07-29