Chaty AI — agentic threat model

Not certain from the listing — likely relies on third-party speech-to-text, LLM, and text-to-speech models. Threats include voice-based prompt injection (vishing-style exploits) and model reprogramming via spoken instructions.

Not certain from the listing — processes real-time availability data, customer PII, and payment details. Threats include unauthorized access to call transcripts, data leakage of customer booking histories, and lack of clear data retention/sanitization policies for voice recordings.

Orchestrates call routing, custom voice generation, and direct API integrations with Rezdy, Roller, Fareharbor, and Bookeo. Threats include insecure tool integration where malicious voice inputs trigger unauthorized API calls, booking modifications, or payment bypasses.

Not certain from the listing — operates on telephony and cloud hosting infrastructure. Threats include SIP trunk hijacking, insecure webhook endpoints connecting to booking platforms, and lack of network isolation between the voice processing unit and internal booking databases.

Provides call transcripts for monitoring. Threats include logging sensitive payment card data (PCI) or PII in plain text within transcripts, and a lack of real-time voice guardrails to detect and block adversarial injection attempts during live calls.

Not certain from the listing — handles booking and payment integrations but does not explicitly state compliance with PCI-DSS, SOC2, or GDPR. Threats include regulatory non-compliance regarding voice recording consent and insecure handling of financial transactions.

Integrates directly with external booking ecosystems (Rezdy, Roller, Fareharbor, Bookeo). Threats include cascading failures if downstream booking APIs are compromised, rate-limiting denial of service, and trust abuse where the agent acts as an authenticated insider within those platforms.