Speechmatics — agentic threat model
Speechmatics acts primarily as a foundational speech-to-text and text-to-speech API rather than an autonomous agent, presenting low direct agentic risk but serving as a critical vector for indirect prompt injection and voice spoofing in downstream applications.
OWASP AIVSS score rationale
| Autonomy of Action | 0.10 | |
| Goal-Driven Planning | 0.00 | |
| Self-Modification | 0.00 | |
| Dynamic Tool Use | 0.10 | |
| Persistent Memory | 0.20 | |
| Contextual Awareness | 0.20 | |
| Dynamic Identity | 0.00 | |
| Multi-Agent Interactions | 0.00 | |
| Non-Determinism | 0.30 | |
| Opacity & Reflexivity | 0.40 |
Scored with the canonical OWASP AIVSS formula (AIVSS calculator reference); agentic risk factors estimated from the agent’s described capabilities.
MAESTRO 7-layer threat model
Per-layer threats for this agent. Layers tagged “not certain from listing” are general, caveated commentary where the public description didn’t pin that layer.
Utilizes proprietary speech-to-text and text-to-speech foundation models. Primary threats include adversarial audio perturbations designed to bypass transcription filters, model extraction/stealing of customizable voice models, and voice cloning/spoofing abuses.
Not certain from the listing — likely processes real-time audio streams and text payloads. Risks include the exposure of sensitive PII within audio transcripts and potential data poisoning if customer audio data is ingested to train or customize voice models.
Not certain from the listing — Speechmatics is an API rather than an orchestration framework. However, downstream agents integrating this API face severe risks of indirect prompt injection if transcribed audio containing malicious instructions is executed blindly by an LLM.
Not certain from the listing — presumably deployed as a high-availability, low-latency cloud API. Key threats include API key theft, lack of transport layer encryption for audio payloads, and denial-of-service attacks targeting real-time processing endpoints.
Not certain from the listing — requires robust logging of API transactions and transcription accuracy monitoring. Gaps in observability could allow silent failures or adversarial manipulation of transcription outputs to go undetected.
Not certain from the listing — processing voice data introduces strict regulatory requirements (e.g., GDPR, CCPA, and biometric data privacy laws). The listing does not specify SOC2, ISO 27001, or specific data retention policies.
Not certain from the listing — acts as a foundational utility within the broader AI ecosystem. Vulnerabilities or outages in this API can cause cascading failures in voice-activated multi-agent systems and customer service bots.
MAESTRO — the 7-layer agentic threat-modeling framework (Cloud Security Alliance / Ken Huang).