Researchers from Mirage Security have uncovered an AI-powered vishing-as-a-service platform, identified as P1 (p1bot.io), which automates “press 1” scams using sophisticated voice technology.

Key Discovery: The P1 Platform

The P1 platform represents a significant shift in vishing (voice phishing) by removing the need for live human scammers during the initial stages of an attack.

• AI Voice Misuse: The platform reportedly misuses text-to-speech (TTS) capabilities from ElevenLabs to generate natural-sounding, AI-driven voices.
• Automation of “Press 1” Scams: Scammers use P1 to spoof phone numbers of trusted institutions, like banks, and play pre-recorded AI messages that claim an account is compromised.
• Live Interaction Removal: Unlike traditional vishing, where a human might need to speak, the AI handles the interaction until a victim is sufficiently “hooked” or directed to a live fraudster for final execution.

Evolving AI Vishing Landscape

The discovery of P1 follows broader academic and industry research into automated vishing:

• ViKing System: In late 2024 and 2025, researchers introduced ViKing, an experimental AI-powered vishing system that used Large Language Models (LLMs) to steer real-time conversations.
• Success Rates: Simulations using these AI systems have successfully extracted sensitive information from up to 77% of participants, as the AI’s fluency often bypasses human suspicion.
• Professionalization: The rise of “vishing-as-a-service” allows low-skill actors to launch high-volume, hyper-personalized attacks for as little as $0.50 to $1.16 per call.

Defensive Measures

As AI makes voice scams nearly indistinguishable from real calls, experts recommend the following:

• Source Verification: Use secondary channels or pre-arranged “code words” to verify a caller’s identity.
• Cognitive Firewalls: Training programs now emphasize “pausing” before acting on urgent requests, as emotional triggers are the primary weapon for AI vishing.
• Real-Time Detection: New tools like isVerified and VishGuard are emerging to identify AI-generated voices during live calls.