Floatbot — agentic threat model

Not certain from the listing — The specific foundation LLMs or speech-to-text/text-to-speech models utilized by Floatbot are not disclosed, leaving potential vulnerabilities to model-specific adversarial prompt injection or evasion attacks unverified.

Not certain from the listing — While Floatbot handles highly sensitive data across banking, healthcare, and insurance domains, the specific architecture for RAG, vector databases, and data isolation between enterprise tenants is not detailed.

Floatbot's 'no-code' AI orchestration framework manages conversational flows and integrates with major contact center technologies (Genesys, NICE, AVAYA, Cisco). Vulnerabilities in this orchestration layer could allow attackers to hijack call routing, manipulate agent-assist prompts, or trigger unauthorized API calls to downstream enterprise systems.

Not certain from the listing — The hosting environment (e.g., public cloud, private cloud, or on-premise deployments) and the sandboxing mechanisms used to isolate execution environments for different enterprise customers are not specified.

Not certain from the listing — Although the platform features an 'AI Coach' to monitor and assist human agents, the internal guardrails, real-time LLM safety monitoring, and logging mechanisms to detect prompt injection or data exfiltration are not described.

Not certain from the listing — Despite claiming 'Enterprise-Grade' status and serving highly regulated industries (HIPAA/PCI-DSS adjacent), the listing does not explicitly confirm specific compliance certifications, identity management standards (SAML/OIDC), or role-based access controls.

Floatbot orchestrates multiple bot types simultaneously (Self-service Chatbots, Voicebots, and AI Agent Assist Bots). This multi-agent ecosystem introduces risks of cascading failures or trust abuse, where a compromised customer-facing voicebot could feed malicious instructions or poisoned context to the internal Agent Assist Bot.