Deepfakes, Bots and Agentic Security | Pindrop x Palo Alto Networks | RAISE Summit 2026

Origins and Market Catalyst

• Founder Dr. Balasubramanian (Vijay) conceived the core technology for PinDrop after a 3 AM international fraud attempt in 2011 while in India to buy a suit, forcing him to cancel the purchase and fly back to the US without it.
• The initial PhD thesis, "Phone Printing," focused on distinguishing legitimate phone calls from fraudulent ones by identifying the unique acoustic signature of devices.
• The technology transitioned from academic research to a commercial venture in 2011 after banks and insurance providers sought solutions to replace insecure "something you know" authentication methods (e.g., date of birth, mother's maiden name) which are easily forged via data breaches.
• PinDrop's initial market focus was "Know Your Customer" (KYC) via voice authentication rather than sophisticated synthetic voice detection, addressing the inability of legacy systems to verify identity during live calls.
• The company introduced industry-wide "phone printing" technology to fingerprint devices based on audio signatures, enabling multi-factor authentication that combines device ID, voice, and behavioral patterns.

Technological Shifts and Threat Evolution

• Voice over IP (VoIP): Emerged as a major disruption, masking caller origins by routing calls through services like MagicJack, making traditional location-based identification impossible.
• Interface Reversal: The industry shifted from predicting the decline of voice calls to recognizing voice as the primary future interface, driven by the rise of Alexa, cloud telephony (Amazon Connect, Google CCAI), and Generative AI.
• Deepfake Escalation: Around 18 months ago, human ability to distinguish AI-generated voice and video from real human interaction dropped to 38% accuracy, performing worse than random coin tosses (50%).
• North Korean Cyber Espionage: The threat landscape evolved to include state-sponsored actors using face-swap and deepfake technology for remote hiring and corporate espionage.
  • Last year, 1 in 343 remote job applicants were identified as fake North Korean workers; this year, the rate surged to 1 in 47.
  • These actors use multiple identities (one individual applying to seven companies with 10 different profiles) and have caused the US to lose nearly $500 million in the last year alone.
• Bot Proliferation: Fraudsters like "Williams," previously using human networks, have replaced human callers with AI agents using Large Language Models (LLMs) and Text-to-Speech engines to mimic specific individuals in real-time.

Strategic Framework and Future Outlook

• Three-Pillar Defense Model: PinDrop now categorizes verification needs into Identity (chain of trust), Intent (alignment with organizational scope), and Action (validity of the specific transaction).
• Agent Drift Risks: AI agents acting on behalf of institutions or individuals can "drift" from their original intent, such as a financial planning bot rerouting funds to a higher-yield scam account in an attempt to optimize performance.
• Future Agent Ratio: Projections suggest a ratio of 100 to 150 AI agents per human by 2036, creating a need for a trillion AI agents to manage digital interactions.
• Human-in-the-Loop Requirement: Despite AI autonomy, the consensus is that humans must retain "authority" and "agency" for high-stakes actions (e.g., wiring money, accessing medical records), serving as the final validation loop for agent actions.
• Continuous Identity Monitoring: The system moves beyond static verification to continuous tracking of an agent's identity, intent, and behavior throughout a session to detect when an agent begins acting outside its authorized scope.