Untapped AI Files Patent: AI Agent Authentication for M2M Security

• Untapped AI, a Cape Town company, has filed a patent with the USPTO for a method allowing AI agents to verify each other's identity during calls.
• This innovation addresses the growing threat of AI voice fraud, where AI can clone voices and facilitate impersonation in machine-to-machine communications.
• The patented system uses an encryption and handshake protocol to ensure fresh authentication for every call, preventing hijacked agents or swarm attacks.
• Unlike existing technologies that only detect a machine's presence, Untapped AI's method verifies *who* the AI agent is and *whom* it represents.
• Untapped AI plans to offer this AI voice fraud prevention patent free to South African companies, emphasizing its importance for AI agent identity verification and machine-to-machine security.

Untapped AI's method provides this critical next step in AI agent authentication.

Untapped AI, a Cape Town-based company specializing in multilingual voice AI solutions for South Africa, has announced the filing of a significant patent application with the United States Patent and Trademark Office (USPTO). The patent describes a novel method designed to enable two artificial intelligence (AI) agents to verify each other's identity during a call before any sensitive information is exchanged. This development is envisioned by Untapped AI's Founder and CEO, Lloyd Matthew, as the crucial next phase in voice communication.

The company anticipates a future where machine-to-machine interactions will dominate phone traffic, with AI systems increasingly communicating through spoken language rather than text. This shift will see personal AI assistants handling calls from various institutional AIs—such as a bank's AI regarding a flagged payment, a clinic's AI rescheduling an appointment, a retailer's AI arranging a product replacement, or an insurer's AI discussing a claim—all without direct human involvement. While such AI assistants are already emerging internationally, and South African businesses have begun deploying their own early voice agents, Untapped AI predicts that machine-to-machine calls will soon surpass human-to-human conversations.

This burgeoning landscape of AI-driven communication, however, inherits and significantly amplifies existing risks associated with phone fraud. South Africa alone experiences billions of rand in annual losses due to fraud, with many cases going unreported. The South African Banking Risk Information Centre (SABRIC) identifies impersonation as a primary method used by fraudsters, a tactic now made more potent by AI's ability to clone voices from brief audio recordings. In a scenario where a fraudster's AI interacts with a consumer's AI, the absence of a human on either end eliminates the natural ability to detect anomalies, making spoofed numbers or borrowed voices far easier to exploit.

Untapped AI's initiative aims to preemptively close this vulnerability before AI-enabled fraud reaches an unmanageable scale. The threat extends beyond simple voice synthesis; it encompasses rogue AI agents that could malfunction, be compromised, be intentionally designed to cause harm, or operate in vast swarms from rotating numbers under falsified personal or business identities. The need for robust AI agent identity verification and Untapped AI machine-to-machine security becomes paramount to mitigate these complex and evolving risks.

Untapped AI's patented solution, an AI voice fraud prevention patent, incorporates an encryption and handshake protocol specifically engineered to counter these sophisticated machine-based threats, drawing on a comprehensive understanding of past and present fraud tactics. The system mandates that every call routed through its network must undergo a fresh authentication process. A failed attempt to prove identity is recorded against the caller across the entire network, not just the specific business called. Any agent unable to verify its identity is refused access and flagged for further investigation.

This design effectively neutralizes swarm attacks, as the handshake is dynamically scrambled and recalibrated for each interaction. Furthermore, a hijacked agent cannot successfully impersonate a legitimate business without passing the rigorous authentication checks. Similar to how humans recognize familiar voices, Untapped AI's method enables two AI agents to recognize a compatible machine through a signal embedded within ordinary-sounding speech, imperceptible to human listeners. Upon successful recognition, the agents establish a separate, encrypted data connection, using the ongoing call to confirm that the party on the data link is indeed the same as the one on the voice line. Only after both sides pass this verification does the interaction transition from spoken conversation to the secure connection, allowing for the safe transmission of sensitive details. Crucially, this confirmation is generated anew for each call, rendering any recorded copy useless for subsequent interactions. Should the other party be a human or a harmless, unverified machine, the agent continues the conversation normally without disclosing sensitive information. However, if impersonation is detected, the agent can terminate the call and trigger an alert. While developers demonstrated in 2025 that AI agents can switch to machine-to-machine sound, and existing patents describe AI callers recognizing each other and moving to data links, Untapped AI's review indicates that these solutions primarily detect *that* a machine is present, not *who* that machine is or *whom* it represents. Untapped AI's method provides this critical next step in AI agent authentication.

This advancement highlights a critical emerging area for legal and compliance professionals, particularly concerning the security of machine-to-machine interactions. The legal implications of AI agent identity are substantial, as businesses will increasingly rely on AI systems to handle sensitive customer data and financial transactions. Ensuring the integrity and authenticity of these AI-driven communications is paramount to prevent fraud, protect data privacy, and maintain regulatory compliance. Lawyers should proactively advise clients on the future necessity of robust AI agent authentication mechanisms to mitigate South Africa AI compliance risk and safeguard against sophisticated cyber threats.

Untapped AI's commitment to offering its USPTO AI agent authentication patent free to South African companies underscores the urgency of addressing these security challenges within the local market. This initiative provides a crucial tool for businesses to enhance their machine-to-machine security protocols, ensuring that AI agents can reliably verify identities before engaging in critical operations. As AI adoption accelerates, the ability to confidently authenticate AI agents will become a cornerstone of secure and trustworthy digital operations, protecting both businesses and consumers from the escalating risks of AI-enabled fraud.

This development highlights a critical emerging area for legal and compliance professionals: securing machine-to-machine interactions. Lawyers should advise clients on the future necessity of AI agent authentication to mitigate fraud and ensure data integrity, especially as solutions like Untapped AI's become available to South African businesses.

Source: Original reporting via ITWeb